Academic motivation, self-concept, engagement, and performance in high school: Key processes from a longitudinal perspective

Abstract

The study tested three theoretically/conceptually hypothesized longitudinal models of academic processes leading to academic performance. Based on a longitudinal sample of 1866 high-school students across two consecutive years of high school (Time 1 and Time 2), the model with the most superior heuristic value demonstrated: (a) academic motivation and self-concept positively predicted attitudes toward school; (b) attitudes toward school positively predicted class participation and homework completion and negatively predicted absenteeism; and (c) class participation and homework completion positively predicted test performance whilst absenteeism negatively predicted test performance. Taken together, these findings provide support for the relevance of the self-system model and, particularly, the importance of examining the dynamic relationships amongst engagement factors of the model. The study highlights implications for educational and psychological theory, measurement, and intervention.

Full text

This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright

Author's personal copy
Academic motivation, self-concept, engagement, and performance in
high school: Key processes from a longitudinal perspective
Jasmine Green
a
, Gregory Arief D. Liem
a
, Andrew J. Martin
a
,
*
, Susan Colmar
a
,
Herbert W. Marsh
b
, Dennis McInerney
c
a
Faculty of Education and Social Work, A35 Education Building, University of Sydney, NSW 2006, Australia
b
Department of Education, University of Oxford, UK
c
Hong Kong Institute of Education, Hong Kong
Keywords:
Motivation
Self-concept
Engagement
Performance
Structural equation modeling
abstract
The study tested three theoretically/conceptually hypothesized longitudinal models of
academic processes leading to academic performance. Based on a longitudinal sample of
1866 high-school students across two consecutive years of high school (Time 1 and Time
2), the model with the most superior heuristic value demonstrated: (a) academic moti-
vation and self-concept positively predicted attitudes toward school; (b) attitudes toward
school positively predicted class participation and homework completion and negatively
predicted absenteeism; and (c) class participation and homework completion positively
predicted test performance whilst absenteeism negatively predicted test performance.
Taken together, these findings provide support for the relevance of the self-system model
and, particularly, the importance of examining the dynamic relationships amongst
engagement factors of the model. The study highlights implications for educational and
psychological theory, measurement, and intervention.
Ó2012 The Foundation for Professionals in Services for Adolescents. Published by Elsevier
Ltd. All rights reserved.
Although academic achievement is important, there has been increasing awareness and empirical interest in psycho-
educational constructs that can be considered key outcomes in education, including motivation (Pintrich, 2003), self-concept
(Marsh, 2007), and engagement (Skinner, Kindermann, Connell, & Wellborn, 2009). In relation to this, theorists (e.g., Pintrich,
2003) have called for a comprehensive and integrative model that reflects the dynamics of different psychoeducational
constructs. Based on the self-system model of motivational development, positing links of context, self, engagement, and
outcomes (Skinner et al., 2009; see also Skinner, Furrer, Marchand, & Kinderman, 2008), the present investigation aims to test
three alternative theoretically/conceptually hypothesized models representing relationships of specific dimensions of the
factors in the model and ascertain their relative effects on performance. Consistent with the posited long-term effects of the
model’s components (Skinner et al., 2009), these hypothesized relationships were tested in longitudinal models that allow
examination of cross-time effects.
The self-system model: a theoretical framework
The self-system model of motivational development (Skinner et al., 2008,2009) posits dynamic relations between
individuals’experience of ‘context’,‘self’,‘engagement/disaffection’, and ‘outcomes’. The notion of self is viewed as indi-
viduals’self-appraisals about their ability and task/activity (e.g., control beliefs, task values) developed through socialization
*Corresponding author. Tel.: þ61 293516273; fax: þ61 2 9351 2606.
E-mail address: andrew.martin@sydney.edu.au (A.J. Martin).
Contents lists available at SciVerse ScienceDirect
Journal of Adolescence
journal homepage: www.elsevier.com/locate/jado
0140-1971/$ –see front matter Ó2012 The Foundation for Professionals in Services for Adolescents. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.adolescence.2012.02.016
Journal of Adolescence 35 (2012) 1111–112 2

Author's personal copy
in a particular context. The model posits that these self-appraisals lead to emotional and behavioral engagement or disaf-
fection.
1
In turn, these patterns of activity (or inactivity) are proposed to impact contextually-relevant outcomes including
achievement and skill acquisitions. The recognition that each component in the model is multidimensional is important to
establish a comprehensive model comprising a comprehensive range of psychoeducational factors.
Skinner et al. (2008,2009) regard engagement (or disaffection) as the central component that not only reflects the
manifestation of motivation and self-related beliefs but also affects outcomes. As reflected in the model, they maintain that
engagement (a) directly predicts learning outcomes, (b) mediates the effects of ‘self’on immediate outcomes, and (c) leads to
subsequent changes in ‘self’, engagement, and outcomes. Effects of each component in the self-system model on its corre-
sponding component over time (feed-forward effects) form a cycle that represents a continuous process of students’moti-
vated engagement with their academic tasks/activities. This cycle explains why students who begin their school academically
engaged become more so, whereas students who start out academically disaffected become gradually more so as they
progress through school (Skinner et al., 2009).
Components and links in the model: context, self, engagement, and outcome
Skinner et al. (2009) acknowledge the multidimensionality of components in the self-system model. However, research
that teases apart these components has been scarce (see, however, Skinner et al., 2008). Thus, the present study aims to
deconstruct components of the model by assessing their specific concepts. In this section, we describe the relevance of the
dimensions selected and hypothesize their conceptual links based on the model and empirical support.
‘Context’in the proposed model: high school
The first component of the model is the ‘context’in which student academic processes are situated. In the present
research, the key context is high school, during which students typically experience significant physical, psychological, and
social changes. While some students are able to navigate this life stage without excessively high levels of turmoil (Eccles et al.,
1993), it is generally recognized that many academic challenges such as diminished motivation, poor self-perceptions, and
disengagement are prevalent during these years. Thus, this period is an important context to investigate.
‘Self’and ‘engagement’in the proposed model: dimensions and inter-relationships
Self: Academic motivation and self-concept. Aligned with the cognitive perspective that recent major motivation theories
are based on (see Pintrich, 2003 for review), the self-system model (Skinner et al., 2009) posits ‘self’as comprising a collection
of beliefs about self in relation to academic ability, task, and activity. Hence, the present study conceptualized the ‘self’
component as belief-based academic motivation (i.e., motivational beliefs) and academic self-concept. Consistent with the
work of Martin (2007,2009), three major types of motivation include adaptive motivation,reflecting an orientation that
facilitates engagement in learning/academic work; impeding motivation, referring to an orientation that inhibits motivated
engagement in learning/academic work; and maladaptive motivation, representing an orientation that is detrimental to
learning/academic work.
Academic self-concept, or students’evaluations of their academic ability, is proposed as another important dimension of
‘self’. Research (see Marsh, 2007 for a review) has shown that academic self-concept is clearly differentiated from general self-
concept (or self-esteem) and that academic self-concept is more highly correlated with academic achievement and behaviors
than are self-esteem and non-academic self-concepts (e.g., social or physical self-concepts). Accordingly, the present study
adopted the domain-specific approach to self-concept by assessing students’academic self-concepts (and not, for example,
their self-esteems).
Engagement: affective and behavior. Academic/school ‘engagement’has been conceptualized as a multidimensional
construct encompassing two or three components (see Appleton, Christenson, & Furlong, 2008). A two-component model
typically comprises affective/emotional and behavioral dimensions (e.g., Finn, 1989;Skinner et al., 2008,2009), whereas
a tripartite model adds cognitive engagement as the third dimension (e.g., Fredricks et al., 2004). The self-system model
explicitly demarcates engagement into affective and behavioral dimensions (Skinner et al., 2008,2009). Indeed, recent
research framed by the model (Skinner et al., 2008) has examined both the external dynamics of the model –which
represents the relations of context, self, action, and outcomes as described above –and its internal dynamics focusing on the
relations of the different engagement dimensions. Importantly, Skinner et al. (2008) have found the salience of emotional
engagement as a predictor of its behavioral counterparts than the other way round. This finding is important in the following
ways. First, it provided support to self-determination (Deci & Ryan, 1985) and effectance motivation (Harter, 1978) theories
1
The self-system model (Skinner et al., 2008,2009) postulates the ‘self’component as encompassing, inter alia, cognition-based constructs developed
through students’continuous interactions with academic tasks and social agents in the school context, such as self-efficacy, self-concept, task value, or goal
orientation. According to multidimensional models of engagement (e.g., Fredricks et al., 2004), these self-belief constructs are examples of key aspects of
cognitive engagement. Hence, the self-system model can be conceptually considered as a model that posits the relational dynamics among dimensions of
cognitive, affective, and behavioral engagement.
J. Green et al. / Journal of Adolescence 35 (2012) 1111–112 21112

Author's personal copy
suggesting that it is emotions (interest, enjoyment) that energize engaged behaviors (effort, school attendance). Second,
consistent with the self-system model (Skinner et al., 2009) and engagement reviewers (e.g., Fredricks et al., 2004), the
finding showed that engagement can be conceptualized as predictor, outcome, or mediator. In view of this, we demarcated
engagement into affective and behavioral dimensions, with the former predicting the latter.
In this study, affective engagement is conceptually represented by positive attitudes toward school which comprises two
distinct but related lower-order concepts: positive school appraisals and positive academic intentions. The former reflects the
extent to which students find schooling an enjoyable and satisfying experience, whereas the latter refers to students’
enthusiasm and interest in continuous involvement in schooling. Behavioral engagement is conceptually represented by
three salient dimensions in student academic behaviors: home-based behavior (homework completion), classroom-based
behavior (classroom participation), and school-relevant behavior (absenteeism). Homework completion reflects the engage-
ment with work set by teachers that students are expected to undertake outside of school hours (Sharp, Keys, & Benefield,
2001). Class participation reflects students’active involvement in the classroom, such as class discussion and group work
during class (Skinner et al., 2009). Absenteeism refers to a period of non-attendance at school (Teasley, 2004) and is typically
the most visible signs of students’behavioral disengagement (Reid, 2005).
2
Motivation and self-concept links to attitudes toward school
Motivation to attitudes. The bulk of evidence has suggested that adaptive motivational beliefs (e.g., self-efficacy, task
values; Bandura, 1997;Wigfield & Eccles, 2000) promote achievement, whereas detrimental motivational beliefs (e.g.,
uncertain control, failure avoidance; Nicholls, 1989;Weiner, 1986) undermine achievement. Further support comes from
a recent review study by Liem and Martin (2012) suggesting that adaptive motivation dimensions are positively associated
with performance whereas impeding and maladaptive motivation dimensions are negatively associated with performance.
Over and above direct effects of motivation on achievement, consistent with the self-system model (Skinner et al., 2008,
2009), the links between motivation and achievement may be mediated by engagement including positive attitudes toward
school. The link between academic motivation and attitudes toward school is supported by a variety of studies. Martin (2007,
2009) has demonstrated relations between motivation and positive academic intentions and positive school appraisals. Other
researchers have shown similar links with a host of other motivational constructs. These include links between self-efficacy
and career/educational aspirations (Bandura, Barbaranelli, Caprara, & Pastorelli, 2001) or enrolment intentions (e.g., Meece,
Wigfield, & Eccles, 1990) and positive links between self-regulation and positive attitudes toward school (Pekrun, Goetz, Titz,
& Perry, 2002).
Self-concept to attitudes. Similar to motivation, Marsh (2007) has demonstrated the direct significance and long-term
effects of academic self-concept on achievement even after controlling for effects of prior achievement. Aligned with the
self-system model (Skinner et al., 2008,2009), however, the relationship between academic self-concept and achievement
may also be mediated by engagement such as students’attitudes toward school or academic subjects. In this regard, the
literature has documented links between academic self-concept and attitudes toward school. For example, Marsh (1991)
investigated the effect of school-average ability on a wide range of educational outcomes including academic aspirations.
When controlling for academic self-concept, the previously confirmed negative effects of school-average ability were reduced
and, as such, academic self-concept showed considerable total and direct effects on academic aspirations. Research has also
shown that self-concept in specific school subjects predicted future course selections (Marsh & Yeung, 1997).
Attitudes links to class participation, homework completion, and absenteeism
Direct effects of positive attitudes toward school or academic subjects on achievement have been empirically demon-
strated (see e.g., Ma & Kishor, 1997 for a meta-analysis). According to the internal dynamics of self-system model (Skinner
et al., 2009), however, effects of affective engagement (e.g., attitudes toward school) on achievement could be mediated
by behavioral engagement. Indeed, as reviewed below, the literature has noted empirical findings suggesting the links of
attitudes toward school to class participation, homework completion, and absenteeism –the three behavioral engagement
dimensions examine here.
Attitudes to class participation. Students with positive orientations to school/learning are more actively engaged in the
class. Valiente, Lemrey-Chalfant, Swanson, and Resier (2008), for example, found that class participation mediated the
relationship linking positive attitudes toward school to grades and absenteeism. Furthermore, research investigating
a school-based intervention showed that increased school satisfaction not only mediated the positive effects of the inter-
vention but also significantly increased participation in physical activity (Dishman et al., 2005).
Attitudes to homework completion. There are mixed findings regarding the link between attitudes toward school and
homework completion. Some research has indicated positive relationships between the two. Trautwein and Köller (2003), for
example, found that students’homework behaviors (e.g., effort and time spent on homework) can be predicted by their
2
It is important to note that school non-attendance can be in the form of unauthorized absenteeism (truancy) or absenteeism with legitimate reasons
(e.g., sickness, family issues, etc.). Thus, absenteeism may not always be an indication of student behavioral disengagement. Regardless of the types of
absenteeism, school non-attendance has been associated with lower academic motivation, engagement, and performance (Reid, 2005).
J. Green et al. / Journal of Adolescence 35 (2012) 1111–1122 1113

Author's personal copy
homework expectancy and academic valuing. Conversely, Cooper (1989) found no significant relationship between pupils’
perceptions of school and their homework completion.
Attitudes to absenteeism. Research has shown that negative attitudes to teachers and low valuing of education are related to
absenteeism (Attawood & Croll, 2006). Research also suggests that school refusal behavior is predicted by students’low
coping expectations and anxiety (see King et al., 2000 for a review of school refusal behavior). Similarly, Vallerand and
colleagues (1997) found that negative intentions about high school predicted subsequent drop-out behavior.
‘Outcome’in the proposed model: factors and inter-relationships
The conceptualization of ‘outcome’in the self-system model encompasses not only achievement but also broader
academic development including skills acquisition (Skinner et al., 2009). Achievement scores represent one form of academic
outcome relevant to the model. Accordingly, students’performance in a standardized test was included in the present study
to represent a key dimension of ‘outcome’in the model.
Behavioral engagement links to performance. The self-system model posits that students’engagement has a direct effect on
their performance (Skinner et al., 2009). Hence, it is reasonable to expect direct links from class participation, homework
completion, and absenteeism to test performance.
Homework completion to performance.InCooper’s (1989) meta-analysis, homework completion was found to positively
predict achievement. However, findings on the relationship between time spent on homework and achievement were
inconclusive; research either demonstrated weak links (Trautwein & KTller, 2003) or no significant relationship (e.g., DeJong,
Westerhof, & Cremmers, 2000). It seems that the amount of homework actually completed by students (rather than the time
spent on homework) is associated with higher achievement (Cooper et al., 1998).
Class participation to performance. Effects of active classroom participation on academic performance are routinely sup-
ported in empirical research (e.g., Valiente et al., 2008). Conversely, a lack of class participation leads to problematic
educational outcomes (Finn, Pannozzo, & Voelkl, 1995) and processes such as emotional withdrawal and poor identification
with school (Finn, 1989), poor academic performance (Finn, 1993), and difficulty following rules and difficulty capitalizing on
learning opportunities (Hughes & Kwok, 2006). Thus, empirical evidence supports a link between class participation and
academic performance.
Absenteeism to performance. School non-attendance has detrimental effects on academic outcomes because absentees
receive less hours of instruction (Rothman, 2002). Supporting evidence shows that absenteeism is a predictor of early school
leaving (Reid, 2005) and poor achievement (e.g., Sutton & Soderstrom, 1999). Similarly, relevant Organisation for Economic
Co-operation and Development (OECD) data found that, in most countries, non-participation (measured by the frequency of
absences, class skipping, and late arrival) had a negative influence on literacy (Willms, 2003). Absenteeism, then, is likely to be
a proximal negative predictor of academic performance.
The present study: a longitudinal approach to the self-system model
Consistent with a call for more comprehensive and integrative models of psychoeducational factors (e.g., Pintrich, 2003),
this investigation adopts the self-system model (Skinner et al., 2009) positing the relations between context, self, engage-
ment/disaffection, and performance. Specifically, the study conceptualized ‘self’through academic motivation and self-
concept;‘engagement’through its affective (positive attitudes toward school) and behavioral (class participation, home-
work completion, absenteeism) dimensions; and ‘outcome’through test performance. In line with the self-system model
(Skinner et al., 2008,2009), our comprehensive reviews of the literature provided support for the relationships of self,
engagement, and outcome factors examined in this study. Thus, we first hypothesized a model reflecting both external and
internal dynamics of the self-system model (Skinner et al., 2009) by specifying the role of emotional engagement in activating
behavioral engagement (Skinner et al., 2008). This model depicts (1) academic motivation and self-concept predicting (2)
attitudes toward school, predicting (3) class participation, homework completion, and absenteeism, predicting (4) test
performance (see Fig. 1a for Hypothesized Model 1).
As parts of a meta-construct of school engagement (Fredricks et al., 2004), emotional and behavioral engagements may be
tightly associated, jointly operative in facilitating learning and achievement, and promoted in similar ways by external factors,
without affecting each other (see Skinner et al., 2008 for a similar view). Thus, with a focus on the external dynamic of the self-
system model (Skinner et al., 2008,2009), we also tested an alternative model hypothesizing (1) academic motivation and self-
concept predicting (2) both emotional and behavioral engagement dimensions (positive attitudes toward school, class
participation, homework completion, and absenteeism), predicting (3) testperformance (see Fig.1b for Hypothesized Model 2).
Furthermore, as reviewed above, research underpinned by motivation and self-concept perspectives has underscored
direct effects on performance of various academic motivational beliefs (Bandura, 1997;Liem & Martin, 2012;Wigfield &
Eccles, 2000) and self-concept (Marsh, 2007;Marsh & Yeung, 1997). Based on these motivation and self-concept theo-
rizing and empirical support, we also tested a third alternative model hypothesizing direct effects of motivation, self-concept,
and engagement on test performance (see Fig. 1c for Hypothesized Model 3).
Aligned with the posited long-term effects of components in the model on their corresponding components (Skinner et al.,
2008,2009), the proposed relationships were examined via three alternative theoretically/conceptually hypothesized
longitudinal models (MacCallum & Austin, 2000). Longitudinal models provide an ideal opportunity to: (a) assess the stability
J. Green et al. / Journal of Adolescence 35 (2012) 1111–112 21114

Author's personal copy
of hypothesized paths over time; (b) examine Time-2 (T2) paths after controlling for Time-1 (T1) variance; and (c) assess
cross-time paths between parallel factors (test-retest paths) in the hypothesized model. Hence, findings derived from each of
these elements will better elucidate the complex dynamics of the self-system model for both each and across time points.
Method
Participants
The longitudinal sample comprises 1866 high school students from six Australian high schools completing the instru-
mentation at T1 (3rd term of the school year) and T2 (one-year later). Approximately 29% of the respondents were in grade 7
a
b
c
Fig. 1. Hypothesized models depicting the predictive relationships of self, engagement and performance. NOTE: AM ¼Adaptive motivation, IM ¼Impeding
motivation, MM ¼Maladaptive motivation, AS ¼Academic self-concept, PA ¼Positive attitudes toward school, PT ¼Class participation, HW ¼Homework,
AB ¼Absenteeism, TP ¼Test performance.
J. Green et al. / Journal of Adolescence 35 (2012) 1111–1122 1115

Author's personal copy
(the beginning of high school) at T1 and grade 8 at T2; 24% were in grade 8 at T1 and grade 9 at T2; 23% were in grade 9 at T1
and grade 10 at T2; 18% were in grade 10 at T1 and grade 11 at T2, and 6% were in grade 11 at T1 and grade 12 (the final year of
high school) at T2. Around 39% of the participants were female. The mean age of respondents was 13.86 years (SD ¼1.28) at T1
and 14.79 years (SD ¼1.28) at T2. This longitudinal dataset is part of a larger Australian Research Council project which has
also investigated academic buoyancy (Martin, Colmar, Davey, & Marsh, 2010) and academic personal bests (Martin & Liem,
2010)–constructs not part of this study.
Measures
Academic motivation. The Motivation and Engagement Scale –High School (MES-HS; Martin, 2007,2009) assesses 11 facets
of motivation which, in the present study, were subsumed into three major dimensions: adaptive motivation, including self-
efficacy (“If I try hard, I believe I can do my schoolwork well”), mastery orientation (“I feel very pleased with myself when I
really understand what I’m taught at school”), valuing of school (“Learning at school is important to me”), persistence (“If I
can’t understand my schoolwork at first, I keep going over it until I understand it”), planning (“Before I start an assignment I
plan out how I am going to do it”), and task management (“When I study, I usually study in places where I can concentrate”);
impeding motivation, including anxiety (“When exams and assignments are coming up, I worry a lot”), failure avoidance
(“Often the main reason I work at school is because I don’t want to disappoint my parents”), and uncertain control (“I’moften
unsure how I can avoid doing poorly at school”), and maladaptive motivation, including self-handicapping (“I sometimes don’t
study very hard before exams so I have an excuse if I don’t do as well as I hoped”) and disengagement (“I often feel like giving
up at school”). Each motivation facet is measured by four items –hence the MES-HS is a 44-item instrument. To each item,
students rate themselves on a 1 (‘Strongly Disagree’)to7(‘Strongly Agree’) scale (see Martin, 2007,2009 for detailed
psychometric properties of the MES-HS). Table 1 present descriptive statistics and reliability of the academic motivation
scales and other scales used in the study for both T1 and T2.
Academic self-concept. To measure students’evaluations of their general academic ability, the four-item academic self-
concept scale drawn from the Self Description Questionnaire II-Short (SDQII-S; Marsh, 1992) was used. An example of the
items is “I am good at most school subjects”. Each SDQ-II-S items are rated on a 1 (‘False’)to6(‘True’) scale.
Engagement. Two four-item scales, positive school appraisals (“I like school”) and positive academic intentions (“I am happy
to stay and complete school”), were used to measure students’positive attitudes toward school. Items are rated on a 1
(‘Strongly Disagree’)to7(‘Strongly Agree’) scale. Homework completion (“How often do you do and complete your homework
and assignments?”) was assessed on a 1 (‘Never’)to5(‘Always’) scale. Class participation (e.g., “I get involved in things we do
in class”) comprises 4 items rated on a 1 (‘Strongly Disagree’)to7(‘Strongly Agree’) scale. Absenteeism (“How many days were
you absent from school last term?”) asked students to specify approximate days absent from school in the previous term.
Responses to the absenteeism item were later coded on a scale of 1 (0 days absent) to 6 (5 or more weeks absent). Homework
completion and absenteeism are single-item measures. The three psychometric scales of positive school appraisal, positive
academic intentions, and class participation were adapted from Martin (2007,2009) who has shown that these scales are
reliable and have a good measurement fit.
Outcome measure. Because class and school grades can be idiosyncratic to classrooms and schools, a standardized
achievement test, the Wide Range Achievement Test-3 (WRAT-3; Wilkinson, 1993), was used to measure academic outcome
measures. Due to time restrictions, this involved the administration of only the spelling and arithmetic subsets (not the
reading subtest) of the WRAT-3. The spelling (untimed) and arithmetic (timed) subtests of the WRAT-3 consist of 40 items
each (administered according to the administration manual, Wilkinson, 1993). Students gained a spelling and arithmetic
performance score based on their performance on the test (standardized by age). As this was a domain-general study (not
a subject-specific study), the spelling and arithmetic performance scores were aggregated to form a general test performance
Table 1
Descriptive statistics, Cronbach’s alphas, and CFA factor loadings for the higher-order factor solution.
Time 1 Time 2
Mean SD Cronbach’s
a
CFA loadings range
(mean)
Mean SD Cronbach’s
a
CFA loadings range
(mean)
AM 5.37 .82 .90 .76–.87 (.81) 5.30 .89 .92 .74–.88 (.82)
IM 3.52 .97 .84 .62–.86 (.70) 3.42 1.06 .85 .60–.84 (.69)
MM 2.47 1.01 .82 .66–.84 (.75) 2.47 1.09 .84 .68–.89 (.79)
AS 4.71 .99 .87 .73–.87 (.80) 4.68 1.05 .87 .72–.87 (.80)
PA 5.52 1.08 .91 .85–.93 (.89) 5.53 1.17 .92 .89–.93 (.91)
PT 5.29 1.20 .91 .78–.90 (.85) 5.26 1.30 .90 .79–.90 (.87)
HW 4.42 .68 –– 4.34 .76 ––
AB 2.14 .84 –– 2.04 .74 ––
TP 109.05 14.01 .67 .64–.79 (.72) 103.93 13.82 .68 .61–.83 (.72)
Note. Homework completion and absenteeism are single-item factors. Test performance is a two-item factor (measured by spelling and arithmetic scores).
AM ¼Adaptive motivation, IM ¼Impeding motivation, MM ¼Maladaptive motivation, AS ¼Academic self-concept, PA ¼Positive attitudes toward school,
PT ¼Class participation, HW ¼Homework, AB ¼Absenteeism, TP ¼Test performance.
J. Green et al. / Journal of Adolescence 35 (2012) 1111–112 21116

Author's personal copy
factor. Two equivalent test forms (blue and tan) for both the spelling and mathematic subtests were utilized at T1 and T2,
respectively.
Procedure
For each school, at each time wave, a classroom teacher was responsible for the administration of the questionnaire in
a normally scheduled class. The teacher first explained the rating scales to students and then presented a sample item.
Students were instructed to complete the instrument independently and to provide only one answer for each item. Spelling
and arithmetic items were presented after completion of the self-report survey items. Once all surveys were collated, the
cover sheet (containing the name of the participant) was discarded and all completed surveys were assigned a unique
identification number that could be used to identify responses for matching T1 and T2 data in longitudinal analyses. This
unique identifier was also used to assure anonymity and confidentiality for all participants.
Statistical analyses
Confirmatory factor analysis and structural equation modeling. The central analyses involved confirmatory factor analysis
(CFA) and structural equation modeling (SEM) using LISREL 8.80 (Jöreskog & Sörbom, 2006). Maximum likelihood was the
method of model estimation used in this research as it is regarded as a robust method with moderate to large sample sizes
(see Hoyle, 1995). Following recommendations on establishing model fit (e.g., Marsh, Hau, & Wen, 2004), the Comparative Fit
Index (CFI), the Non-Normed Fit Index (NNFI), the Root Mean Square Error of Approximation (RMSEA), the
c
2
test statistic,
and an evaluation of parameter estimates were used in the current research to assess model fit. RMSEA values at or less than
.05 are taken to reflect excellent fit(Yuan, 2005). The NNFI and CFI varyalong a 0-to-1 continuum in which values at or greater
than .95 are typically taken to reflect excellent fit(McDonald & Marsh, 1990).
Treatment of missing data. The Expectation-Maximization (EM) algorithm has been recommended as an appropriate
missing data method that minimizes biased parameter estimates and inaccurate standard errors (Schafer & Graham, 2002).
The percentage of missing data for this study was relatively low at 4.88% at T1 and 4.82% at T2; hence, the EM algorithm was
considered appropriate to handle missing data.
Correlated uniquenesses. In order to obtain precise estimates of relations among corresponding latent constructs at each
time wave, correlations among their uniquenesses must be included in the model (Marsh, Roche, Pajares, & Miller, 1997). If
correlated uniquenesses are not included in longitudinal models, the relations between the latent constructs (i.e., parameter
estimates) may be positively biased (Marsh et al., 1997). Accordingly, all longitudinal CFAs and SEMs include correlated
measurement errors for parallel T1 and T2 items.
Results
Longitudinal confirmatory factor analysis
We first examined factor structures of the measures by conducting a longitudinal CFA comprising 18 factors (9 factors for
each time wave): four higher-order factors (adaptive, impeding, and maladaptive motivation and attitudes toward school)
and five first-order factors (self-concept, class participation, homework completion, absenteeism, and test performance). This
measurement model fit the data well (
c
2
¼36,974.70, df ¼7,922, RMSEA ¼.06, CFI ¼.96, NNFI ¼.96). The summary of factor
loadings (all significant at p<.001) are reported in Table 1. Further, multi-group testing demonstrated gender and year-level
(junior, middle, senior high school) invariance of the longitudinal CFA (RMSEA ¼.04, CFI ¼.96, NNFI ¼.96), providing
justification for conducting whole-sample analyses. Table 2 shows that factors predicted to be related in the model were
significantly correlated, providing support to testing hypothesized relationships in each of the three hypothesized models
that take into account shared variances among constructs.
Longitudinal structural equation modeling
Model 1. Having shown that psychometric properties of the measures were sound and robust, the analysis was then
focused on testing hypothesized models. The first hypothesized model (Fig. 1a) comprised: (1) motivation and self-concept
predicting (2) attitudes toward school, predicting (3) class participation, homework completion, and absenteeism, predicting
(4) test performance. This model showed an excellent fit to the data (
c
2
¼23,991.29, df ¼7,876, p<.001, RMSEA ¼.04,
CFI ¼.98, NNFI ¼.98). All the hypothesized paths were significant in the predicted directions (positive or negative). The
results also indicated congruence of predictive paths across the two time waves, demonstrating the stability of the
hypothesized model over time. That is, all T2 paths remained significant even after controlling for shared variance with the
parallel T1 factors (see Fig. 2). Furthermore, T1 factors positively predicted their parallel T2 factors (test-retest paths), sup-
porting the posited long-term effects of factors in the model over time (Skinner et al., 2009).
Model 2. The second hypothesized model (Fig. 1b) depicted (1) motivation and self-concept predicting (2) attitudes toward
school, class participation, homework completion, and absenteeism, predicting (3) test performance. This model showed an
excellent fit to the data (
c
2
¼23,771.70, df ¼7,850, p<.001, RMSEA ¼.04, CFI ¼.98, NNFI ¼.98). However, several
J. Green et al. / Journal of Adolescence 35 (2012) 1111–1122 1117

Author's personal copy
hypothesized paths in this model were not significant. For example, the path from impeding motivation to class participation
(
b
¼.02 at T1;
b
¼.02 at T2) and absenteeism (
b
¼.06 at T1;
b
¼.05 at T2). Interestingly, at Time 1, the path from class
participation to test performance became negative (
b
¼.08, p<.05). Given that the T1 correlation between class partici-
pation and test performance was positive (r¼.14 , p<.001), this negative regression weight may indicate a suppression effect
due to the shared variance in test performance jointly explained by class participation and the other three predictors,
especially positive attitudes toward school. That is, the significant correlation between class participation and positive
attitudes toward school (r¼.55, p<.001) –which was relatively higher than the correlations between class participation and
homework completion (r¼.28, p<.001) and absenteeism (r¼.07, p<.05) –may have led to multicollinearity. Of note, the
fact that class participation and positive attitudes toward school shared around 30% of variance may also point to their
conceptual overlap under the broad construct of school engagement, with the former being a behavioral indicator and the
latter being an affective indicator of school engagement (Fredricks et al., 2004).
Model 3. The third hypothesized model (Fig. 1c) depicted motivation, self-concept, and all the engagement factors (atti-
tudes toward school, class participation, homework completion, absenteeism) as direct predictors of test performance. This
model also showed a good fit to the data (
c
2
¼25,386.99, df ¼7,852, p<.001, RMSEA ¼.04, CFI ¼.98, NNFI ¼.98). Similar to
Model 2, however, several paths in this model were not significant. For example, the path from impeding motivation to test
performance at T1 (
b
¼.01) and one from class participation to test performance at T2 (
b
¼.03).
Model comparisons. Although our data seemed to fit the three hypothesized models well,
c
2
statistics suggested that Model
1(
Dc
2
¼1395.70,
D
df ¼24, p<.001) and Model 2 (
Dc
2
¼1615.70,
D
df ¼2, p<.001) had a better fit than Model 3, suggesting
that the two models hypothesized according tothe self-system model (Skinner et al., 2008,2009) were statistically superior to
Model 3 hypothesized based on eclectic perspectives of self, motivation, and engagement. Moreover,
c
2
statistics also indicated
Table 2
Within- and between-time longitudinal higher-order CFA correlations.
AM IM MM AS PA PT HW AB TP
AM .65 .28/.29 .84/.84 .54/.56 .77/.78 .53/.58 .51/.45 .15/.19 .17/.27
IM .18/.22 .80 .60/.52 .59/.52 .28/.26 .28/.28 .26/.21 .04/.05 .18/.12
MM .55/.55 .31/.37 .75 .57/.60 .78/.79 .52/.56 .57/.51 .15/.20 .24/.28
AS .37/.37 .45/.41 .43/.42 .62 .56/.58 .46/.50 .44/.39 .12/.13 .34/.32
PA .48/.46 .19/.17 .48/.51 .38/.38 .63 .55/.60 .42/.39 .17/.25 .39/.42
PT .37/.34 .17/.20 .52/.36 .30/.36 .35/.33 .59 .28/.25 .07/.09 .14/.14
HW .36/.35 .17/.16 .41/.40 .27/.33 .32/.23 .17/.18 .52 .16/.20 .31/.26
AB .08/.09 .02/.02 .08/.12 .08/.05 .16/.13 .07/.02 .12/.07 .34 .20/.24
TP .15/.20 .13/.11 .19/.23 .31/.26 .34/.35 .16/.09 .18/.28 .17/.17 .87
Note.AM¼Adaptive motivation, IM ¼Impeding motivation, MM ¼Maladaptive motivation, AS ¼Academic self-concept, PA ¼Positive attitudes toward
school, PT ¼Class participation, HW ¼Homework, AB ¼Absenteeism, TP ¼Test performance. Within-Time 1/within-Time 2 correlations are reported in
upper diagonal, whereas between-time (Time 1-Time 2/Time 2-Time 1) correlations are reported in lower diagonal. Test-retest correlations are bolded in
diagonal. All rvalues >.04 are significant at p<.05.
Fig. 2. Final Standardized Structural Relations in the Longitudinal Model 1. NOTE: *p<.05, **p<.01, ***p<.001; AM ¼Adaptive motivation, IM ¼Impeding
motivation, MM ¼Maladaptive motivation, AS ¼Academic self-concept, PA ¼Positive attitudes toward school, PT ¼Class participation, HW ¼Homework,
AB ¼Absenteeism, TP ¼test performance. Time 1–Time 2 test-retest paths are indicated in brackets [ ]. Fit indices:
c
2
¼23,991.29, df ¼7876, RMSEA ¼.04,
CFI ¼.98, NNFI ¼.98.
J. Green et al. / Journal of Adolescence 35 (2012) 1111–112 21118

Author's personal copy
that Model 2 had a better fit than Model 1 (
Dc
2
¼219.59,
D
df ¼26, p<.001). Given a number of key hypothesized paths in
Model 2 did not reach statistical significance, model comparison indicated a higher heuristic value of Model 1 over Model 2.
Furthermore, the apparent suppression effect observed in Model 2 highlights the importance of conceptualizing affective
engagement as a predictor of behavioral engagement as in Model 1. Hence, model comparisons provide support for the
heuristic superiority of Model 1, in which the internal dynamic of the self-system model (Skinner et al., 2008) was specified.
Supplementary analyses
The above analyses have provided empirical support for Models 1 and 2 developed based on the self-system model. Hence,
we further examined modified Models 1 and 2 by testing the salience of T1 motivation and self-concept in predicting T2
engagement and achievement (see Fig. 3). These models had a good fit to the data:
c
2
¼7050.88, df ¼1,925, p<.001,
RMSEA ¼.04, CFI ¼.98, NNFI ¼.98 (modified Model 1) and
c
2
¼7137.29, df ¼1913, p<.001, RMSEA ¼.04, CFI ¼.98,
NNFI ¼.97 (modified Model 2). Results also indicated a generally similar pattern of predictive paths between the main models
and their respective modified models. As shown in Fig. 3a, all the paths in the modified Model 1 were significant except one
path from adaptive motivation to positive attitude toward school (
b
¼.11, ns; cf. Fig. 2). As shown in Fig. 3b, although most of
the paths in the modified Model 2 were found to be in the hypothesized directions, except the negative path from class
participation to test performance (
b
¼.16, p<.001), many of the key paths in the modified Model 2 were not statistically
significant (e.g., all the predictive paths from adaptive motivation) –which were congruent with the initial Model 2. Taken
together, although our supplementary analyses provided support for the self-system model (Skinner et al., 2008,2009),
consistent with our main findings, they underscored the heuristic superiority of Model 1 over Model 2.
Discussion
Juxtaposing three alternative models, our analyses demonstrated that the two models hypothesized based on the self-
system framework (Skinner et al., 2008,2009) had a statistically better fit than the model built upon eclectic perspectives
of self, motivation, and engagement (Fredricks et al., 2004;Marsh, 2007). Of the two better fitting models, however, the model
that conceptualized affective engagement as a predictor of behavioral engagement (Model 1) showed a more superior
heuristic value over the model that positioned affective and behavioral engagement as concurrent predictors of achievement
(Model 2). Specifically, Model 1 is one depicting (a) academic motivation and self-concept predicted attitudes toward school;
(b) attitudes toward school positively predicted class participation and homework completion and negatively predicted
absenteeism; and (c) class participation and homework completion positively predicted test performance whilst absenteeism
negatively predicted test performance. Importantly, in this model, most predictive paths at T2 remained significant after
controlling for shared variance with T1 counterparts, demonstrating the stability of the hypothesized model over time.
Further, cross-time effects showed that T1 factors positively predicted their corresponding T2 factors –consistent with
Skinner et al. (2009) positing the continuous impacts of components in the model over time, leading to either a virtuous cycle,
Fig. 3. Final standardized structural relations in the modified longitudinal models 1 and 2. NOTE: *p<.05, **p<.01, ***p<.001; AM ¼Adaptive motivation,
IM ¼Impeding motivation, MM ¼Maladaptive motivation, AS ¼Academic self-concept, PA ¼Positive attitudes toward school, PT ¼Class participation,
HW ¼Homework, AB ¼Absenteeism, TP ¼Test performance. Fit indices for modified Model 1:
c
2
¼7050.88, df ¼1,925, RMSEA ¼.04, CFI ¼.98, NNFI ¼.98; Fit
indices for modified Model 2:
c
2
¼7137.29, df ¼1,913, RMSEA ¼.04, CFI ¼.98, NNFI ¼.97.
J. Green et al. / Journal of Adolescence 35 (2012) 1111–1122 1119

Author's personal copy
when motivated and engaged students are progressively more so, or a vicious cycle when disaffected students become more
detached to their schoolwork over time. Taken together, the findings highlight the relevance of the self-system model and the
importance to specifically examine the dynamic relationships amongst engagement factors (Skinner et al., 2008).
Implications for theory and research
The study holds important implications for educational theory and research. First, the study has shown that key
components of the self-system model can be ‘unpacked’to their specific dimensions. The distinction and hypothesized
ordering of affective and behavioral dimensions not only affirms the multidimensional perspective of engagement (Fredricks
et al., 2004), but also provides evidence to the ‘internal’dynamics of the self-system model that postulates the role of affects
in the manifestation of engaged behaviors (Skinner et al., 2008). Further, the differential effects of class participation,
homework completion, and absenteeism on academic outcomes shows the fruitfulness of using these specific dimensions
resembling three conceptual categories of behavioral engagement proposed by Fredricks et al. (2004): positive conduct
(including school attendance), involvement with learning, and participation in school activities. Of note, this demonstrates
that some behavioral factors are more critical and more adaptive than others.
Second, the study offers new perspectives on self-concept and motivation research. As shown here, when motivation and
self-concept are considered in one model, not only do they share variance (evidenced through their significant correlation),
both are also effective in explaining unique variance in criterion variables (evidenced through their respective significant
betas). These findings demonstrate a complementary and synergic role of motivation and self-concept in student academic
trajectory, particularly in their school engagement. Furthermore, the reciprocal effect model (Marsh, 2007) postulates that,
over and above test-retest paths, prior self-concept should predict subsequent performance and prior performance should
predict subsequent self-concept. Inspired by these findings, the logical extension of the reciprocal model should focus on
testing the mediated effect of motivation in the reciprocal relation between self-concept and performance and, further, test
the synergic role of motivation and self-concept by determining their interaction effects on engagement.
Intervention opportunities
The constructs under focus in the self-system model (particularly academic motivation and self-concept) have a theo-
retical basis which provided a direction for intervention. The self-system model (Skinner et al., 2009) is broadly based on
a view that components of the self-system process model are not fixed stable ‘traits’; rather, they reflect constructs relatively
malleable and open to intervention by teachers, parents, and students themselves. Hence, enhanced engagement can be
expected from interventions that target improvements of academic motivation and self-concept (see O’Mara, Marsh, Craven,
& Debus, 2006).
Limitations, future directions, and conclusion
There are limitations to consider when interpreting findings and which form a direction for future research. First, our
findings suggested the heuristic superiority of Model 1, highlighting the stability of the relationships amongst constructs
under consideration at two time points with a one-year interval and, importantly, the robustness of these relationships at T2
after controlling for shared variance with T1 factors. However, the key predictive parameters (
b
s) in Model 1 were estimated
based on the relationships of self, engagement, and performance factors within the same time point (cf. modified Model 1 in
which T1 self factors predicted T2 engagement factors). Given the lack of temporal precedence between predictors and
outcomes, the observed predictive paths in Model 1 should be interpreted with caution. Second, although our final model was
built upon theoretical/conceptual perspectives, future studies should test other alternative models, such as one conceptu-
alizing prior engagement and achievement as predictors of subsequent motivation and self-concept. Third, the data are
predominantly self-reported (though standardized spelling and arithmetic achievement data were collected). Findings could
be further illuminated through the use of additional measures such as teacher/parent reports (e.g., for engagement) and class/
school records (e.g., for absenteeism). Fourth, correlations of key factors in this study were relatively high and this may have
lead to multicollinearity which has affected their joint contributions on performance, particularly in Models 2 and 3. Future
research should measure other factors that have better discriminant validity (i.e., lower correlations) as predictors of
engagement and achievement. Fifth, the study was conducted from a domain-general perspective and did not consider its
domain-specific nature. The literature has documented support for the utility of domain specificity of various motivational
constructs (e.g., Green, Martin, & Marsh, 2007) and the present model might be expanded to consider this. We believe that
these research directions are important to provide further evidence for the self-system model.
References
Appleton, J. J., Christenson, S. L., & Furlong, M. J. (2008). Student engagement with school: critical conceptual and methodological issues of the construct.
Psychology in Schools, 45, 369–386.
Attawood, G., & Croll, P. (2006). Truancy in secondary school pupils: prevalence, trajectories, and pupil perspectives. Research Papers in Education, 21, 467–484.
Bandura, A. (1997). Self-efficacy: The exercise of control. New York: W.H. Freeman.
J. Green et al. / Journal of Adolescence 35 (2012) 1111–112 2112 0

Author's personal copy
Bandura, A., Barbaranelli, C., Caprara, G. V., & Pastorelli, C. (2001). Self-Efficacy beliefs as shapers of children’s aspirations and career trajectories. Child
Development, 72,187–206.
Cooper, H. (1989). Homework. White Plains, NY: Longman.
Cooper, H., Lindsay, J. L., Nye, B., & Greathouse, S. (1998). Relationships among attitudes about homework, amount of homework assigned and completed,
and student achievement. Journal of Educational Psychology, 90,70–83.
Deci, E. L., & Ryan, R. M. (1985). Intrinsic motivation and self-determination in human behavior. New York: Plenum.
DeJong, R., Westerhof, K., & Cremmers, B. (2000). Homework and student math achievement in junior high schools. Educational Research and Evaluation, 6,
130–157.
Dishman, R. K., Motl, R. W., Saunders, R., Felton, G., Ward, D., Dowda, M., et al. (2005). Enjoyment mediates effects of a school-based physical activity
intervention. Medicine and Science in Sports and Exercise, 37,478–487.
Eccles, J. S., Wigfield, A., Midgley, C., Reuman, D., Mac Iver, D., & Feldlaufer, J. (1993). Negative effects of traditional middle schools on students’motivation.
The Elementary School Journal, 93, 553–574.
Finn, J. D. (1989). Withdrawing from school. Review of Educational Research, 59,117–142.
Finn, J. D. (1993). School engagement and students at risk. Washington, DC: U.S Department of Education, National Center of Educational Studies.
Finn, J. D., Pannozzo, G. M., & Voelkl, K. E. (1995). Disruptive and inattentive-withdrawn behaviour and achievement among fourth graders. The Elementary
School Journal, 95,421–434.
Fredricks, J. A., Blumenfeld, P. C., & Paris, A. H. (2004). School engagement: potential of the concept, state of the evidence. Review of Educational Research, 74,
59–90.
Green, J., Martin, A. J., & Marsh, H. W. (2007). Motivation and engagement in English, mathematics and science high school subjects: towards an under-
standing of multidimensional domain specificity. Learning and Individual Differences, 17, 269–279.
Harter, S. (1978). Effectance motivation reconsidered: toward a developmental model. Human Development, 21,36–64.
Hoyle, R. H. (1995). The structural equation modelling approach: basic concepts and fundamental issues. In R. H. Hoyle (Ed.), Structural equation modelling:
Concepts, issues, and applications (pp. 1–15). CA: Sage Publications Thousand Oaks.
Hughes, J. N., & Kwok, O. (2006). Classroom engagement mediates the effect of teacher-student support on elementary students’peer acceptance:
a prospective analysis. Journal of School Psychology, 43, 465–480.
Jöreskog, K. G., & Sörbom, D. (2006). LISREL 8.80 for Windows [Computer Software]. Lincolnwood, IL: Scientific Software International, Inc.
King, N. J., Tonge, B. J., Heyne, D., & Ollendick, T. H. (2000). Research on cognitive-behavioral treatment of school refusing children: a review and
recommendations. Child Psychology Review, 20, 495–507.
Liem, G. A. D., & Martin, A. J. (2012). The Motivation and engagement scale: theoretical framework, psychometric properties, and applied yields. Australian
Psychologist, 47,3–13 .
Ma, X., & Kishor, N. (1997). Assessing the relationship between attitude toward mathematics and achievement in mathematics: a meta-analysis. Journal for
Research in Mathematics Education, 28,27–47.
MacCallum, R. C., & Austin, J. T. (2000). Applications of structural equation modelling in psychological research. Annual Review of Psychology, 51,201–226.
McDonald, R. P., & Marsh, H. W. (1990). Choosing a multivariate model: noncentrality and goodness of fit. Psychological Bulletin, 107,247–255.
Marsh, H. W. (1991). Failure of high ability schools to deliver academic benefits commensurate with their students’ability levels. American Educational
Research Journal, 28,445–480.
Marsh, H. W. (1992). Self-Description Questionnaire (SDQ) III: A theoretical and empirical basis for the measurement of multiple dimensions of late adolescent
self-concept: A test manual and a research monograph. Publication Unit, SELF Research Centre, University of Western Sydney.
Marsh, H. W. (2007). Self-concept theory, measurement and research into practice: The role of self-concept in educational psychology. Leicester, UK: British
Psychological Society.
Marsh, H. W., Hau, K. T., & Wen, Z. (2004). In search of golden rules: comment on hypothesis testing approaches to setting cutoff values for fit indexes and
dangers in overgeneralising Hu & Bentler’s (1999) findings. Structural Equation Modelling, 11, 320–341.
Marsh, H. W., Roche, L. A., Pajares, F., & Miller, D. (1997). Item-specificefficacy judgments in mathematical problem solving: the downside of standing too
close to trees in a forest. Contemporary Educational Psychology, 22, 363–377.
Marsh, H. W., & Yeung, A. S. (1997). The causal effects of academic self-concept on academic achievement: structural equation models of longitudinal data.
Journal of Educational Psychology, 89,41–54.
Martin, A. J. (2007). Examining a multidimensional model of student motivation and engagement using a construct validation approach. British Journal of
Educational Psychology, 77,413–440.
Martin, A. J. (2009). Motivation and engagement across the academic lifespan: a developmental construct validity study of elementary school, high school,
and university/college students. Educational and Psychological Measurement, 69, 794–824.
Martin, A. J., Colmar, S. H., Davey, L. A., & Marsh, H. W. (2010). Longitudinal modeling of academic buoyancy and motivation: do the ‘5Cs’hold up over time?
British Journal of Educational Psychology, 80,473–496.
Martin, A. J., & Liem, G. A. D. (2010). Academic Personal Bests (PBs), engagement, and achievement: a cross-lagged panel analysis. Learning and Individual
Differences, 20, 265–270.
Meece, J. L., Wigfield, A., & Eccles, J. S. (1990). Predictors of math anxiety and its influence on young adolescents’course enrolment intentions and
performance in mathematics. Journal of Educational Psychology, 82,60–70.
Nicholls, J. G. (1989). The competitive ethos and democratic education. Cambridge: Harvard University Press.
O’Mara, A. J., Marsh, H. W., Craven, R. G., & Debus, R. L. (2006). Do self-concept interventions make a difference? A synergistic blend of construct validation
and meta-analysis. Educational Psychologist, 41,181–206.
Pekrun, R., Goetz, T., Titz, W., & Perry, R. P. (2002). Positive emotions in education. In E. Frydenberg (Ed.), Beyond coping: Meeting goals, visions, and challenges
(pp. 149–174). Oxford, UK: Elsvier.
Pintrich, P. R. (2003). Motivation and classroom learning. InReynolds, W. M., Miller, G. E., & Weiner, I. B. (Eds.). (2003). Handbook of psychology: Educational
psychology, Vol. 7 (pp. 103–122). Hoboken, NJ: John Wiley & Sons.
Reid, K. (2005). The causes, views, and traits of school absenteeism and truancy: an analytical review. Research in Education, 74,59–82.
Rothman, S. (2002). Student absence in South Australian schools. Australian Educational Researcher, 29,69–91.
Schafer, J. L., & Graham, J. W. (2002). Missing data: our view of the state of the art. Psychological Methods, 7,147–17 7.
Sharp, C., Keys, W., & Benefield, P. (2001). Homework: A review of recent research. Slough: National Foundation for Educational Research (NFER).
Skinner, E. A., Furrer, C., Marchand, G., & Kinderman, T. (2008). Engagement and disaffection in the classroom: Part of a larger motivational dynamic? Journal
of Educational Psychology, 100,765–781.
Skinner, E. A., Kindermann, T. A., Connell, J. P., & Wellborn, J. G. (2009). Engagement as an organizational construct in the dynamics of motivational
development. In K. Wentzel, & A. Wigfield (Eds.), Handbook of motivation in school (pp. 223–245). Malwah, NJ: Erlbaum.
Sutton, A., & Soderstrom, I. (1999). Predicting elementary and secondary school achievement with school-related and demographic factors. The Journal of
Educational Research, 92,330–338.
Teasley, M. L. (2004). Absenteeism and truancy: risk, protection, and best practice implications for school social workers. Children and Schools, 26,117–128.
Trautwein, U., & KTller,O. (20 03). The relationship between homeworkand achievement –still much of a mystery. Educational Psychology Review,15,115–145.
Valiente, C., Lemrey-Chalfant, K., Swanson, J., & Resier, M. (2008). Prediction of children’s academic competence from their effortful control, relationships,
and classroom participation. Journal of Educational Psychology, 100,67–77.
Vallerand, R. J., Fortier, M. S., & Guay, F. (1997). Self-determination and persistence in a real-life setting: toward a motivation model of high school drop.
Journal of Personality and Social Psychology, 72,1161–1176 .
J. Green et al. / Journal of Adolescence 35 (2012) 1111–1122 1121

Author's personal copy
Weiner, B. (1986). An attributional theory of motivation and emotion. New York: Springer-Verlag.
Wigfield, A., & Eccles, J. S. (2000). Expectancy-value theory of achievement motivation. Contemporary Educational Psychology, 25,68–81.
Wilkinson, G. (1993). WRAT-3: Wide Range Achievement Test administration manual. Wilmington, DE: Wide Range Inc.
Willms, J. D. (2003). Student engagement at school: A sense of belonging and participation. Results from PISA 2000. Paris: Organisation for Economic
Co-operation and Development (OECD).
Yuan, K. H. (2005). Fit indices versus test statistics. Multivariate Behavioural Research, 40,115–148.
J. Green et al. / Journal of Adolescence 35 (2012) 1111–112 2112 2