Putwain2022 (PEDIR POR SEXO)

Journal of Educational Psychology
© 2020 American Psychological Association 2022, Vol. 114, No. 1, 108–126

ISSN: 0022-0663 http://dx.doi.org/10.1037/edu0000637
Achievement Emotions and Academic Achievement: Reciprocal Relations

and the Moderating Influence of Academic Buoyancy
David W. Putwain and Peter Wood Reinhard Pekrun

Liverpool John Moores University University of Essex and Australian Catholic University
Control-value theory proposes that achievement emotions impact achievement, and that achievement
outcomes (i.e., success and failure) reciprocally influence the development of achievement emotions.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Academic buoyancy is an adaptive response to minor academic adversity, and might, therefore, offer
This document is copyrighted by the American Psychological Association or one of its allied publishers.
protection from achievement being undermined by negative achievement emotions. At present, however,
there is little empirical evidence for these hypothesized relations. In this study we examined reciprocal
relations between three achievement emotions (enjoyment, boredom, and anxiety) and test performance
in the context of mathematics, and whether academic buoyancy moderated relations between these
emotions and test performance. Data were collected from 1,242 primary school students (mean age ⫽ 9.3
years) over 4 waves within 1 school year. Achievement emotions (T1 and T3) and test performance (T2
and T4) were measured alternately. Academic buoyancy was measured at T3. A structural equation model
showed negative relations of anxiety to subsequent test performance and negative relations of test
performance to subsequent anxiety. Test performance also predicted enjoyment and boredom, but not
vice versa. A latent-interaction structural equation model showed buoyancy moderated relations between
anxiety and test performance. Test performance was highest when anxiety was low and buoyancy high.
Practitioners should consider using interventions to reduce anxiety and downstream effects on
achievement.
Educational Impact and Implications Statement

In classroom settings, multiple emotions such as enjoyment, boredom, and anxiety may occur.
Among these emotions, anxiety is especially important for students’ achievement in mathematics
according to this study with elementary schoolchildren. Reducing anxiety would be beneficial for
students’ achievement, for example through fostering their adaptive responses to failure and increas-
ing perceptions of control. The findings also suggest that developing academic buoyancy can benefit
the achievement of students with mild forms of anxiety.
Keywords: academic achievement, achievement emotions, anxiety, buoyancy, control-value theory
Mathematics learning in elementary or primary school is gen- nushek & Woessmann, 2008). Furthermore, economic national
erally considered to be of critical importance to the person and for competitiveness requires a highly skilled science, technology, en-
society at large. Functional numeracy skills are vitally important in gineering, and mathematics (STEM) workforce (Kärkkäinen &
adult personal and work life (e.g., managing personal finances), Vincent-Lancrin, 2013). Unfortunately, many students fail to learn
and failure to master basic mathematics skills is associated with these fundamental numeracy skills. For example, in the United
subsequent unemployment and lower earning potential (e.g., Ha- States, 19% of children were judged to show below basic mathe-
matics skills at the end of Grade 4 in 2019 (National Center for
Educational Statistics, 2019). Similarly, 21% of children left pri-
This article was published Online First October 8, 2020. mary school in England at the age of 11 years in 2019 without
X David W. Putwain and X Peter Wood, School of Education, Liver- reaching the expected minimum standard in mathematics (Depart-
pool John Moores University; X Reinhard Pekrun, Department of Psy- ment for Education, 2019).
chology, University of Essex, and Institute of Positive Psychology and Achievement emotions have been found to be critically impor-
Education, Australian Catholic University. tant for students’ academic achievement (e.g., Loderer, Pekrun, &
We have no known conflict of interest to disclose. Our work was funded Lester, 2018; Tze, Daniels, & Klassen, 2016; von der Embse,
by a grant from the Bowland Trust. We thank Emma Rainbird for her
Jester, Roy, & Post, 2018). However, few studies have been
assistance with data collection.
Correspondence concerning this article should be addressed to David W. conducted that take into consideration the combined influence of
Putwain, School of Education, Liverpool John Moores University, IM multiple achievement emotions. Furthermore, the impact of pos-
Marsh Campus, Mossley Hill Road, Liverpool L17 6BD, United Kingdom. sible protective factors has been neglected. The present longitudi-
E-mail: [email protected] nal study with Year 5 primary schoolchildren targets these deficits
108
EMOTIONS, ACHIEVEMENT, AND BUOYANCY 109
by examining joint relations between three emotions that students ment may also depend on interactions of these mechanisms with
commonly experience in the classroom, namely enjoyment, bore- task demands. Specifically, effects on achievement mediated by
dom, and anxiety, and their performance in mathematics. In addi- different styles of processing may depend on the match between
tion, we investigated whether an asset-driven psychological attri- processing style and type of task (see, e.g., Fiedler & Beier, 2014).
bute, namely academic buoyancy, is influential in protecting For example, as proposed in CVT, anxiety can facilitate rigid
students’ mathematics test performance from the detrimental in- rehearsal of learning materials, whereas enjoyment can enhance
fluences of boredom and anxiety. more creative ways of studying, implying that these two emotions
can promote different kinds of task performance. Given the inter-
Achievement Emotions play between mediating mechanisms and their interactions with
type of task, the effects of emotions on students’ learning may be
complex. However, for resulting academic achievement, it is rea-
Concept of Achievement Emotions
sonable to expect that greater enjoyment typically results in greater
Achievement emotions were defined by Pekrun (2017) as “emo- achievement. In turn, academic success would, all things being
tions that relate to achievement activities (e.g., participating in a equal, strengthen control and positive value appraisals resulting in
competition) or achievement outcomes (e.g., success and failure)” greater enjoyment (a positive reciprocal cycle; for supporting
(p. 252). Achievement emotions are multifaceted, containing af- evidence, see Pekrun, Lichtenfeld, Marsh, Murayama, & Goetz,
fective, cognitive, physiological, and motivational components, 2017).
and are distinct from moods which are of lower intensity, longer Boredom. In CVT boredom is theorized to impair perfor-
lasting, and often with no specific referent (Linnenbrink, 2006; mance and achievement by undermining interest and intrinsic
Pekrun, 2006). Achievement emotions can be experienced in a motivation, reducing cognitive resources, and promoting superfi-
variety of achievement-oriented settings, including the classroom, cial learning. Types of boredom characterized by very low arousal
tests and exams, and homework. In the present study we focused (indifferent and apathetic) may be more damaging for learning
on three classroom-related emotions: enjoyment, boredom, and than those (e.g., searching and reactant) characterized by an active
anxiety. These emotions can be differentiated by valence and search for less boring alternatives (Goetz & Hall, 2014). Students
physiological activation. Enjoyment is a positive activating emo- in English Year 5 primary school mathematics lessons do not have
tion, boredom is a negative deactivating emotion, and anxiety is a the option to choose alternative (and potentially less boring) ac-
negative activating emotion. tivities. It is likely that the more damaging types of boredom
(indifferent and apathetic) would be experienced. Lower perfor-
Relations Between Achievement Emotions and mance and success would, all things being equal, weaken value
appraisals, resulting in greater boredom (a positive reciprocal
Academic Achievement: Control-Value Theory
cycle).
Control-value theory (CVT; Pekrun, 2006, 2018, in press; Anxiety. In CVT anxiety is theorized to have dual effects on
Pekrun & Perry, 2014; Pekrun, Goetz, Titz, & Perry, 2002, 2019) performance and achievement. Interest and intrinsic motivation
is a theoretical framework that incorporates the antecedents and can be undermined and working memory processes and executive
outcomes of achievement emotions. Appraisals of value (how and functions disrupted (also see Eysenck & Derakshan, 2011; Ey-
why an achievement activity or outcome is important) and control senck, Derakshan, Santos, & Calvo, 2007). However, anxiety can
(expectancy for future, and attributions of past success and failure) also facilitate more rigid information processing, such as simple
are considered as proximal antecedents of achievement emotions. rehearsal, and can increase motivation to invest effort to avoid
Achievement emotions are thought to influence cognitive and failure. Rigid information processing is unlikely to benefit students
motivational processes that, in turn, underpin performance and on the mathematics test used in the present study as the items
academic achievement. According to CVT, not only will achieve- required not only the recall of previously learnt mathematical
ment emotions influence performance and achievement, but per- reasoning but the application of that reasoning to a novel question.
formance and achievement outcomes can also influence achieve- Although increased effort will likely reduce the negative effects of
ment emotions in a cycle of reciprocal causation. anxiety, the overall effects of anxiety on academic achievement are
Enjoyment. In CVT, enjoyment is theorized to enhance aca- negative in most students; a recent meta-analysis of 238 studies
demic achievement though promoting interest and intrinsic moti- showed a mean correlation of r ⫽ ⫺.24 between test anxiety and
vation, maintaining cognitive resources, focusing attention on the achievement (von der Embse et al., 2018). Lower performance and
task at hand, and supporting use of flexible and deep learning success would, all things being equal, weaken control and negative
strategies as well as self-regulation of learning. These different value appraisals resulting in greater anxiety (a positive reciprocal
motivational and cognitive mechanisms can interact. For example, cycle).
the influence of positive affect on cognition and attention might
differ according to the degree of motivational intensity (Gable & Empirical Studies of Relations Between Achievement
Harmon-Jones, 2010). When positive affect is combined with less
Emotions and Academic Achievement
intense approach motivation (e.g., in contentment), cognition and
attention can be broadened, but when combined with high- When considered in isolation, enjoyment shows positive, and
intensity approach motivation (e.g., in enjoyment), cognition and boredom and anxiety show negative, relations with academic
attention can be narrowed to focus on the most salient task details achievement (for meta-analyses see Loderer et al., 2018; Tze et al.,
to facilitate goal pursuit. As such, enjoyment should be especially 2016; von der Embse et al., 2018). Few studies have modeled
conducive to academic learning. Furthermore, effects on achieve- enjoyment, boredom, and anxiety together, however, to account
110 PUTWAIN, WOOD, AND PEKRUN
for the concurrent relations between these three emotions and their enjoyment and achievement (␤s .12 to .30), and boredom and
unique contributions to predicting achievement. Three notable achievement (␤s ⫺.07 to ⫺.36); enjoyment and boredom were
exceptions have included enjoyment, boredom, and anxiety, to- shown to have unique effects. No studies thus far have included
gether in single analytic models. In secondary school students, more than two emotions simultaneously in a single analytic model.
Ahmed, van der Werf, Kuyper, and Minnaert (2013) found unique In the present study we address this limitation of the literature.
statistically significant relations with mathematics achievement for
enjoyment (positively) and boredom and anxiety (negatively). Academic Buoyancy
However, Raccanello, Brondino, Moe, Stupnisky, and Lichtenfeld
(2019; elementary school students) and Putwain, Schmitz, Wood,
What Is it, and How Does It Differ From
and Pekrun (2020; primary school students) found that only en-
joyment and anxiety, but not boredom, remained statistically sig-
Cognate Constructs?
nificant unique predictors of subsequent mathematics achieve- Academic buoyancy is the ability to respond adaptively to the
ment. everyday challenges, setbacks, and pressures experienced by stu-
A few studies that have included enjoyment, boredom, or anx- dents during their studies (Martin & Marsh, 2009). Examples are
iety, alongside other emotions not included in the present study, periods of poor performance; dips in confidence, motivation, and
further underscore how the predictive value of discrete emotions engagement; receiving negative feedback from teachers; and the
can differ when considered together with other emotions. When demands of tests and assessments. Academic buoyancy can be
considered alongside pride, enjoyment positively and anxiety neg- contrasted with academic resilience which refers to adaptive re-
atively predicted mathematics achievement in undergraduates sponses to major adversities such as chronic underachievement
(Villavicencio, & Bernardo, 2016). Neither enjoyment nor bore- and failure, school truancy and refusal, and clinical levels of
dom or anxiety were significantly related to the mathematics anxiety or depression (Martin & Marsh, 2009). In short, academic
achievement of elementary school students when considered buoyancy is the ability to bounce back from minor adversities,
alongside surprise, curiosity, confusion, and frustration (Muis, whereas academic resilience is the ability to bounce back from
Psaradellis, Lajoie, Di Leo, & Chevrier, 2015). Another study of major adversities.
elementary school students that included the same achievement Fong and Kim (2019) showed that academic buoyancy was
and epistemic emotions found that boredom negatively predicted distinct from other cognate constructs, including grit (i.e., persis-
mathematics achievement, mediated by lower use of cognitive tency of effort and consistency of interest; Duckworth, Peterson,
learning strategies (Di Leo et al., 2019). Matthews, & Kelly, 2007) and future time perspective (i.e., the
Achievement emotions are often moderately to highly intercor- perceived connection between present activities and future goals;
related (e.g., rs ⫽ ⫺.62 to .82 in Pekrun, Goetz, Frenzel, Barch- Lens & Seginer, 2015), in a sample of undergraduate students.
feld, & Perry, 2011) because of the shared appraisal antecedents Items for academic buoyancy loaded separately from those of grit
(Pekrun, 2006). The colinearity between emotions reduces the and future time perspective in factor analysis, and correlations
predictive power for each discrete emotion; only those with stron- between academic buoyancy and the other two constructs were
ger relations to performance and achievement, or that have rela- small (rs ⬍ .27).
tively weaker colinearity with other emotions, will remain statis- An unresolved question in the literature is the extent to which
tically significant predictors. The practical corollary in classroom some level of exposure to adversity is necessary for persons to
settings is that when achievement emotions co-occur, not all may build adaptive responses (e.g., Brooks, 2006; Compas, 2004). The
exert effects of the same strength on performance and achieve- types of adversities that academic buoyancy is theorized to protect
ment; some emotions may be more practically significant for against are those experienced during routine schooling by the
performance than others. Therefore, it is vital for empirical work to majority of students (Martin & Marsh, 2009). Indeed the very
combine achievement emotions to competitively identify the most utility of the academic buoyancy construct is founded on this
robust predictors of performance and achievement. point; unlike resilience it has relevance to the majority of students.
Evidence has shown that children in primary or elementary school
Reciprocal Relations Between Achievement Emotions can experience and overcome difficulties in reading, writing, and
and Academic Achievement numeracy (e.g., Holmes & Dowker, 2013; O’Connor, Beach, San-
chez, Bocian, & Flynn, 2015), may receive negative feedback from
Studies testing reciprocal relations between achievement emo- teachers (Hattie & Timperley, 2007), negatively compare them-
tions and achievement are largely lacking, with three notable selves to higher achieving classmates (Marsh, 2007), and may be
exceptions. Over 10 waves of measurement reciprocal relations exposed to the pressures of testing (e.g., von der Embse & Witmer,
were shown between boredom and test performance (␤s ⫺.10 2014). These are the types of everyday academic adversities that
to ⫺.23) in undergraduate students taking an introductory psychol- buoyancy is theorized to protect against, and they are adequately
ogy course (Pekrun, Hall, Goetz, & Perry, 2014). Over five waves captured by the academic buoyancy scale (Martin & Marsh, 2008).
of measurement, Pekrun et al. (2017) showed reciprocal relations
with mathematics achievement for enjoyment (␤s .11 to .13),
Relations With Achievement
boredom (␤s ⫺.06 to ⫺.09), and anxiety (␤s ⫺.07 to ⫺.14) in
secondary school students. Only one study (Putwain, Becker, et al., Although academic buoyancy shows positive relations with
2018; primary school students) has modeled reciprocal effects for adaptive beliefs, affect, and behaviors in primary and secondary
more than a single emotion (enjoyment and boredom) simultane- school students (e.g., Hirvonen, Yli-Kivistö, et al., 2019; Hir-
ously. Over four waves, reciprocal relations were shown between vonen, Putwain, et al., 2019; Martin, Colmar, Davey, & Marsh,
2010, 2013; Putwain, Connors, Symes, & Douglas-Osborn, 2012), The Buffering Effect of Academic Buoyancy for
relations between academic buoyancy and achievement are equiv- Adaptive Educational Outcomes
ocal. Putwain, Daly, Chamberlain, and Saddredini (2016) showed
that domain general academic buoyancy positively predicted per- Academic buoyancy is linked to adaptive responses to adversity,
including strengthened positive and reduced negative emotions. In
formance in aggregated scores for English, science, and mathe-
addition, it is plausible that academic buoyancy would not only
matics secondary school exit examinations (␤ ⫽ .16), after con-
lessen the intensity of emotions such as anxiety and boredom but
trolling for concurrent relations with test anxiety. Martin (2014)
also reduce their educational detrimental impact. Low levels of
showed that greater academic buoyancy predicted higher achieve-
academic buoyancy would be expected to have little impact on the
ment (␤ ⫽ .07) on standardized secondary school numeracy and
negative relations between boredom and anxiety, on the one hand,
literacy tests, after controlling for sociodemographical variables and achievement, on the other. As buoyancy increases, it would be
and Big Five personality traits. expected to buffer against the detrimental impact of boredom and
However, other studies have shown that buoyancy did not anxiety such that the negative relation would be weakened. An
always predict achievement when included in more complex mod-
interaction between academic buoyancy and boredom or anxiety

els with multiple predictors. In studies of secondary school stu-
would therefore be expected. At low boredom or anxiety, there

dents that have included control, academic buoyancy did not would be little difference in the achievement of low and high
predict academic achievement on standardized school numeracy academically buoyant students. As boredom and anxiety increase,
and literacy tests (Collie, Martin, Malmberg, Hall, & Ginns, 2015) however, high academically buoyant students will show higher
or secondary school exit examinations in English, mathematics, achievement that their low academically buoyant counterparts.
and science (Putwain & Aveyard, 2018). In the aforementioned Few studies have examined the possible moderating role of
study by Fong and Kim (2019), academic buoyancy was not academic buoyancy. Putwain, Daly, Chamberlain, and Sadreddini
significantly related with self-reported GPA in undergraduates (2015) showed that the negative relation between test anxiety and
after controlling for grit and future time perspective. task-focused coping was reduced in academically buoyant second-
Although the bivariate correlations in these studies (Collie et al., ary school students. Symes, Putwain, and Remedios (2015) found
2015; Fong & Kim, 2019; Putwain & Aveyard, 2018) were posi- that the negative relation between teachers’ use of failure-
tive (rs ⫽ .10 to .15), in the presence of related variables the avoidance messages prior to a forthcoming high-stakes exam, on
predictive value of academic buoyancy was reduced. This reduc- the one hand, and the appraisal of that exam as threatening, on the
tion of direct effects may be attributable to the effects of buoyancy other, was reduced in academically buoyant secondary school
being mediated by other variables (e.g., buoyancy bolstering per- students. Finally, Martin and Marsh (2019) found a marginally
ceived control, and control influencing performance in the studies significant effect (␤ ⫽ ⫺.13, p ⫽ .10) of the interaction between
by Collie et al., 2015, and Putwain & Aveyard, 2018). Alterna- academic buoyancy and academic adversity on subsequent aca-
tively, the reason may be construct overlap reducing the individual demic adversity in a sample of secondary school students. The
predictive power of buoyancy when combined with, for instance, positive relation between prior and later academic adversity, a year
apart, was weaker at higher academic buoyancy in keeping with its
the perseverance component of grit. Furthermore, reduced direct
theorized adaptive nature.
effects do not rule out the possibility that buoyancy interacts with
No studies have examined how academic buoyancy buffers the
other variables, especially those that pose academic adversity
effects of performance impairing negative classroom emotions,
(such as low perceived control). It is also possible that small or
such as boredom and anxiety. In the present study we address this
statistically nonsignificant relations between academic buoyancy
gap in the literature. Because low enjoyment is not typically
and achievement also arise from using domain-general measures considered as academically adverse, we would not expect enjoy-
of academic buoyancy and achievement or mismatching domain- ment to interact with academic buoyancy but keep this as an
general measures of academic buoyancy with domain-specific exploratory question.
measures of achievement (see Swann, Chang-Schneider, & Larsen
McClarty, 2007).
Only two studies have used domain-specific measures of aca- Aims and Hypotheses
demic buoyancy to examine relations with achievement. Yun,
Given the age of the participants in our study, namely students
Hiver, and Al-Hoorie (2018) showed that academic buoyancy in
in Year 5 in their penultimate year of primary education (Year 5)
second language (L2) acquisition predicted end-of-course L2 exam aged 9 –10 years, we chose to measure classroom emotions spe-
scores (␤ ⫽ .31) after controlling for prior achievement in a cifically. In England, where there study was based, students take
sample of undergraduate students. Colmar, Liem, Connor, and standardized National Curriculum Tests (NCTs) in reading (two
Martin (2019), however, found only small relations between aca- papers) and mathematics (two papers) at the end of Year 2, aged
demic buoyancy for mathematics and mathematics test perfor- 6 –7 years (Key Stage 1 NCTs), and in English (three papers) and
mance (r ⫽ .10), and between academic buoyancy for reading mathematics (three papers) at the end of primary schooling, Year
and reading test performance (r ⫽ .09), in primary school 6, aged 10 –11 years (Key Stage 2 NCTs). Key Stage 1 NCTs are
students. Yun et al.’s (2018) findings suggest that there is some administered informally and marked by teachers whereas Key
merit in the idea that relations between academic buoyancy and Stage 2 NCTs are administered formally and marked by an exter-
achievement are stronger when analyzed in a domain-specific nal agency.
fashion. Accordingly, in the present study we adopted a Students in Year 5, therefore, have fewer experiences of formal
domain-specific approach. testing than in other countries (notably the United States) and have
less homework (or self-study) than students in secondary or higher in presecondary education. Second, we examined whether aca-
education. To capture the typical affective learning experiences of demic buoyancy moderated relations between achievement emo-
English students at this age it is preferable to focus on classroom tions and test performance. The second novel contribution of our
experiences. We focused on enjoyment, boredom, and anxiety study was, therefore, to test whether academic buoyancy could
partly because they are three of the most commonly experienced protect achievement from emotions like boredom and anxiety.
achievement emotions (see Pekrun et al., 2002). Furthermore, the Like with achievement emotions, to ensure a robust evidence base
only instrument presently available for measuring achievement for the buoyancy construct, it is essential for empirical work to use
emotions in elementary/primary schoolchildren (Lichtenfeld, younger students in primary (or elementary) schooling as well as
Pekrun, Stupnisky, Reiss, & Murayama, 2012) contains scales for older students in secondary or university education. Only one
these three emotions. When combined with the need to limit the study to date (Colmar et al., 2019) has examined academic buoy-
number of items required by participating schools, and use high- ancy in primary school students. The present study, therefore,
quality age-appropriate instruments, we took the pragmatic deci- makes a noteworthy contribution to the understanding of academic
sion to focus on enjoyment, boredom, and anxiety. buoyancy by also using a sample of students in primary education.
The aim of the study was twofold. First, we examined reciprocal Succinctly stated, we tested the following hypotheses (see Fig-
relations between achievement emotions and achievement, with ure 1):

multiple emotions entered simultaneously in the same analytical
Hypothesis 1: Enjoyment and academic buoyancy are posi-
model. More specifically, we aimed to test a model of reciprocal
tively related, and boredom and anxiety are negatively related,
relations between three classroom achievement emotions (enjoy-
to subsequent test performance.
ment, boredom, and anxiety) and test performance in a sample
of Year 5 primary school students (aged 9 –10 years) in the domain Hypothesis 2: Test performance is positively related to sub-
of mathematics, over four waves. We alternated the assessment of sequent enjoyment and negatively related to subsequent bore-
emotions and test performance over waves as sequential models dom and anxiety.
are suited to test reciprocal relations (Little, Preacher, Selig, &
Hypothesis 3: Academic buoyancy attenuates the negative
Card, 2007; Pekrun et al., 2014; Rosel & Plewis, 2008). Thus the
relations between boredom and subsequent test performance
first novel contribution of this study was to examine enjoyment,
and between anxiety and subsequent test performance
boredom, and anxiety, simultaneously using a four-wave design in
a sample of primary school students. Despite the (often) greater
Method
ethical and logistical challenges involved, to ensure a robust and
generalizable evidence base for CVT is it critical that empirical Participants and Procedure
work uses samples of presecondary students as well as those in
secondary and higher education. The present study contributes to Data were collected over four waves; self-report data at T1 and
the paucity of empirical studies to use samples of younger students T3 and mathematics test performance at T2 and T4 (see Figure 1).
T1 Enjoyment T3 Enjoyment
T1 Boredom T2 Maths Test T3 Boredom T4 Maths Test
T1 Anxiety T3 Anxiety
Figure 1. The hypothesized fully forward reciprocal model including autoregressive, cross-lagged, and
concurrent relations between achievement emotions and mathematics test performance. Paths for gender and age
are not depicted.
All data were collected in the participants’ classrooms at school by test were less likely to participate at T3, t(977) ⫽ 4.87, p ⬍ .001,
the regular classroom teacher following a standardized script. The and T4, t(734) ⫽ 6.58, p ⬍ .001. All other differences were not
survey items and mathematics tests were hosted online and statistically significant (ps ⬎ .05). Because missing data could be
prompted students where they had missed an answer. This was to accounted for by T2 mathematics test performance, they were
minimize missing data arising from participants inadvertently treated as missing at random (MAR) and handled using full-
missing an item. T1 and T2 data were collected in December 2018, information-maximum-likelihood (FIML) estimation. FIML has
and T3 and T4 data in June 2019. Mathematics tests were sched- been found to result in trustworthy, unbiased estimates for MAR
uled for approximately one week after the surveys. The project was when the variable causing missingness is included in the model
approved by an institutional research ethics committee (19/EHC/ (Nicholson, Deboeck, & Howard, 2017), even in the case of a high
01) at the first and second authors’ university. Written consent was amount of missing values (Enders, 2010), and to be an adequate
provided by the head teacher of each participating school and the method to manage missing data in longitudinal studies (Jeličič,
parent or carer of each participating student. Individual verbal Phelps, & Lerner, 2009).
assent for each participant was sought at each wave of data
collection. The script required teachers to check the voluntary

Measures
participation of each student, verbally, and provide an alternative

activity if the students declined to participate. Students were in- Emotions. Achievement emotions were measured at T1 and
formed that all survey and test responses would remain anonymous T3, using the 12 items from the Achievement Emotions
and not be seen by teachers or parents. These points were also Questionnaire-Elementary School classroom emotions scales
explained on the online survey and mathematics tests. Although (AEQ-ES: Lichtenfeld et al., 2012). These scales measure three
timed, the classroom setting and anonymity of results would char- achievement emotions (enjoyment, boredom, and anxiety) experi-
acterize these tests as being low-stakes. enced in classroom settings with four items each (e.g., “I look
At T1, data were collected from 1,242 students (633 male, 609 forward to maths lessons” for enjoyment; “I find maths lessons so
female; mean age ⫽ 9.3 years, SD ⫽ .49) from 24 English primary boring I would rather do something else” for boredom; “When I
schools (45 different classrooms). The ethnic backgrounds of think about maths lessons, I get nervous” for anxiety). All items
participants were Asian ⫽ 246 (19.8%), black ⫽ 58 (4.7%), were mathematics-specific and adapted to use parlance typical for
white ⫽ 876 (70.5%), Chinese ⫽ 11 (0.9%), other ⫽ 22 (1.8%), the English context (e.g., “class” changed to “lesson”). Participants
and mixed heritage ⫽ 29 (2.3%). Data could not be collected for responded to items on a 5-point scale (1 ⫽ not at all, 5 ⫽ very
the socioeconomic backgrounds of individual students due to much). Internal consistency coefficients for the present study were
private data protection reasons. However, the schools were located excellent (see Table 1).
in two of 12 nationally designated opportunity areas characterized Buoyancy. Academic buoyancy was measured at T3 using the
by relatively high levels of deprivation (Department for Education, Academic Buoyancy Scale (ABS; Martin & Marsh, 2008). The
2017). ABS comprises four items that were adapted to be mathematics-
There was attrition at subsequent waves of data collection (T2 specific and the example included in the one item simplified to
n ⫽ 979, T3 n ⫽ 863, and T4 n ⫽ 734) resulting from participants make appropriate to the age of the target sample (“I’m good at
either being absent from school at the time of data collection or dealing with setbacks in maths at school [e.g., getting a question
from choosing not to participate. To assess potential bias in miss- wrong]”). Participants responded to these items on a five-point
ing data we conducted an omnibus test for missing completely at scale (1 ⫽ not at all, 5 ⫽ very much). The internal consistency of
random (MCAR; Little’s test, Little, 1988). This was followed by the scale in the present study was excellent (see Table 1).
a series of t tests comparing mean values of T1 and T3 age and Performance. Mathematics test performance was measured at
emotions, and T2 and T4 mathematics test performance, as well as T2 and T4 using items pooled from the 2016, 2017, and 2018 Key
logistic regressions for T1 and T3 frequencies of gender, for Stage 2 National Curriculum Test (NCT) reasoning papers (Stan-
participants with complete versus incomplete data. Little’s test was dards and Testing Agency, 2016a, 2016b, 2017a, 2017b, 2018a,
statistically significant (p ⬍ .001) indicating that MCAR could not 2018b). NCTs are tests taken by English schoolchildren at the end
be assumed. Participants who scored lower on the T2 mathematics of primary schooling (Year 6) covering the curriculum taught from
Table 1
Descriptive Statistics and Item Factor Loadings
Item M SD ␣/␻ ␳I Skewness Kurtosis Factor loadings
T1 Enjoyment 16.24 4.48 .92/.92 .07 ⫺1.09 0.21 .84–.90

T1 Boredom 7.87 4.38 .91/.91 .04 1.12 0.16 .79–.88
T1 Anxiety 8.30 4.38 .82/.82 .03 0.99 0.17 .70–.75
T3 Enjoyment 15.40 4.75 .93/.93 .06 ⫺0.86 ⫺0.27 .84–.96
T3 Boredom 7.78 4.38 .91/.92 .04 0.94 ⫺0.11 .86–.89
T3 Anxiety 8.23 4.27 .82/.83 .04 0.99 0.24 .79–.90
T3 Buoyancy 15.58 3.85 .79/.79 .03 ⫺0.51 ⫺0.28 .64–.78
T2 Mathematics test performance 4.64 3.67 .79/.81 .12 0.82 0.15 —
T4 Mathematics test performance 9.55 4.72 .85/.85 .14 0.11 ⫺0.73 —
Note. ␻ ⫽ McDonald’s omega; ␳I ⫽ intraclass correlation coefficient (ICC1).
Years 3 to 6 (Key Stage 2; age 7 to 11 years). There are three All latent models were estimated using maximum-likelihood with
mathematics NCTs: one 30-min arithmetic paper and two 40-minte robust standard errors (MLR) and evaluated using the following
reasoning papers scheduled over two days using a paper-and- model fit indices: Root mean error of approximation (RMSEA),
pencil format. Each paper consisted of a series of closed-response standardized root mean residual (SRMR), confirmatory fit index
questions worth between one and three marks each that used (CFI), and Tucker-Lewis index (TLI). A good model fit is indicated
constructive and substantive styles of reasoning (Bohn-Gettler, by RMSEA ⬍.05, SRMR ⬍.06, and CFI and TLI ⬎.95 (Hu &
2019; Forgas, 2008). Questions required convergent analytical Bentler, 1999). Caution must be used, however, when applying guid-
thinking (reasoning with one correct solution) rather than diver- ance derived from simulations studies to more complex studies, such
gent and more creative thinking. The maximum score was 40 for as ours, using naturalistic data (Heene, Hilbert, Draxler, Ziegler, &
the arithmetic paper and 35 for each of the reasoning papers (the Bühner, 2011; Lance, Butts, & Michels, 2006).
exact number of questions differs in each paper depending on the
number of marks allocated to each question).1
Results
All schools in England follow a prescribed national curriculum
which, for mathematics, at Key Stage 2 covers arithmetic, mea-

surement, geometry, fractions and statistics, ratio and proportion, Descriptive Statistics
and simple algebra (Department for Education, 2013). As partic-
Participants reported relatively high levels of enjoyment and
ipants in the present study were in Year 5, we did not include
academic buoyancy and low levels of boredom and anxiety (see
arithmetic questions (typically taught in Years 3 and 4) because
Table 1). T1 enjoyment showed a slightly negative and T1 boredom
these are easier and may have resulted in a ceiling effect. We asked
a slightly positive skew (hence the decision to use the MLR
two primary school mathematics teachers unrelated to this study to
estimator). The internal consistency of self-report measures (Cron-
select all those items from the 2016, 2017, and 2018 reasoning
bach’s ␣s and McDonald’s ␻s ⱖ .79) and mathematics tests (␣s ⱖ
papers that would be appropriate for a Year 5 student (questions on
.79 and ␻s ⱖ.81) were good, and items loaded substantively on
measurement, geometry, fractions and statistics, were included).
their target factors (␭s ⱖ .64) in CFA (see Table 1). Intraclass
The resulting pool of items was subsequently confirmed as being
correlation coefficients (ICC1 or ␳I) showed that the proportion of
appropriate for Year 5 students by two mathematics learning
variance attributable to the school level was relatively small for the
specialists unconnected to the present study.
classroom achievement emotions and academic buoyancy (approx-
Items were randomly selected from this pool to create two tests.
imately 3%–7%) and somewhat larger for the mathematics test
Each test was timed for 40 min (to correspond to that of a NCT
performance (approximately 12%–14%).
reasoning paper) and contained one- and two-point questions
worth 20 marks in total (the first test comprised of 17 questions
and the second test 18 questions). Responses required students to Latent Bivariate Correlations
either provide a numerical value, or choose one or more answers
from a list or menu of options. Unlike NCTs, marks were awarded To estimate latent bivariate correlations, a measurement model
for a correct answer only and no marks were awarded for correct was built that included achievement emotions (four items each for
reasoning when an incorrect answer was given. An exemplar item enjoyment, boredom, and anxiety at T1 and T3), academic buoy-
is: “A box contains 2.6 kg of washing powder. Jack used 65 grams ancy (four items at T3), and mathematics test performance at T2
of powder per wash. He uses all of the powder. How many washes and T4. The corresponding indicators for classroom achievement
did Jack do?” Participants were not provided with feedback on emotions at T1 and T3 were allowed to correlate. Mathematics test
their test score. The internal consistency for the two tests was performance was treated as a single-item latent variable with ␴ε ⫽
excellent (see Table 1). .1 in line with estimates derived from previous empirical studies
Demographic variables. We controlled for self-reported gen- (Hoy, Tarter, & Hoy, 2006; Watkins, Lei, & Canivez, 2007).
der (0 ⫽ male, 1 ⫽ female) and age in the analysis. Gender (0 ⫽ male, 1 ⫽ female) and age were added as covariates
and treated as manifest variables.
The CFA showed a good fit to the data, ␹2(298) ⫽ 589.30, p ⬍
Data Analysis .001, RMSEA ⫽ .026, SRMR ⫽ .028, CFI ⫽ .982, and TLI ⫽
.977, and so we proceeded to inspect correlation coefficients (see
A latent variable modeling approach was adapted using Mplus
Table 2). Enjoyment and buoyancy were positively, and boredom
v.8 (Muthén & Muthén, 2017). Confirmatory factor analysis
and anxiety negatively, correlated with mathematics test perfor-
(CFA) was employed to check the properties of measurement
mance. Female students reported lower enjoyment, higher anxiety
models, check for measurement invariance, and estimate latent
and boredom, lower buoyancy, and lower mathematics test per-
bivariate correlations. Structural equation modeling (SEM) was
formance. Older students reported lower enjoyment and higher
used to estimate reciprocal relations between emotions and
boredom, and they showed higher mathematics test performance.
achievement, and a latent interaction structural equation model
The intercorrelations between T1 emotions (rs ⫽ ⫺.68 to .64)
(LI-SEM) was used to estimate the moderating effects of academic
buoyancy on the relations between emotions and achievement. The and T3 emotions and academic buoyancy (rs ⫽ ⫺.78 to .62)
type ⫽ complex command was used to adjust standard errors for indicate the possibility of multicollinearity in subsequent analyses
the clustering of data. Using single-level modeling and adjusting that model constructs simultaneously. T4 mathematics test perfor-
standard errors is recommended to account for nestedness when
relations between variables at higher levels are not the target of 1
Papers and mark schemes can be found at https://www.gov.uk/
investigation (Wu & Kwok, 2012). government/collections/national-curriculum-assessments-practice-materials.
Table 2
Correlations Between the Study Variables
Variable 1 2 3 4 5 6 7 8 9 10 11
ⴱⴱ ⴱⴱⴱ ⴱⴱⴱ ⴱⴱⴱ ⴱⴱⴱ ⴱⴱⴱ ⴱⴱ ⴱⴱ ⴱⴱⴱ
1. T1 Enjoyment — ⫺.68 ⫺.36 .60 ⫺.48 ⫺.24 .44 .14 .13 ⫺.17 ⫺.07ⴱ
2. T1 Boredom ⫺.64ⴱⴱⴱ — .64ⴱⴱⴱ ⫺.40ⴱⴱⴱ .45ⴱⴱⴱ .28ⴱⴱⴱ ⫺.28ⴱⴱⴱ ⫺.18ⴱⴱⴱ ⫺.17ⴱⴱⴱ .07ⴱ .11ⴱⴱⴱ
3. T1 Anxiety ⫺.33ⴱⴱⴱ .56ⴱⴱⴱ — ⫺.24ⴱⴱⴱ .25ⴱⴱⴱ .51ⴱⴱⴱ ⫺.38ⴱⴱⴱ ⫺.32ⴱⴱⴱ ⫺.36ⴱⴱⴱ .12ⴱⴱ .01
4. T3 Enjoyment .57ⴱⴱⴱ ⫺.41ⴱⴱⴱ ⫺.27ⴱⴱⴱ — ⫺.78ⴱⴱⴱ ⫺.40ⴱⴱⴱ .62ⴱⴱⴱ .18ⴱⴱⴱ .16ⴱⴱ ⫺.15ⴱⴱⴱ ⫺.01
5. T3 Boredom ⫺.48ⴱⴱⴱ .56ⴱⴱⴱ .27ⴱⴱⴱ ⫺.75ⴱⴱⴱ — .58ⴱⴱⴱ ⫺.43ⴱⴱⴱ ⫺.19ⴱⴱⴱ ⫺.19ⴱⴱⴱ .11ⴱⴱ .05
6. T3 Anxiety ⫺.27ⴱⴱⴱ .31ⴱⴱⴱ .45ⴱⴱⴱ ⫺.38ⴱⴱⴱ .53ⴱⴱⴱ — ⫺.47ⴱⴱⴱ ⫺.34ⴱⴱⴱ ⫺.40ⴱⴱⴱ .09ⴱ .01
7. T3 Academic buoyancy .37ⴱⴱⴱ ⫺.26ⴱⴱⴱ ⫺.33ⴱⴱⴱ .52ⴱⴱⴱ ⫺.28ⴱⴱⴱ ⫺.36ⴱⴱⴱ — .25ⴱⴱⴱ .23ⴱⴱⴱ ⫺.10ⴱⴱ .02
8. T2 Test performance .15ⴱⴱⴱ ⫺.17ⴱⴱⴱ ⫺.29ⴱⴱⴱ .21ⴱⴱⴱ ⫺.21ⴱⴱⴱ ⫺.32ⴱⴱⴱ .22ⴱⴱⴱ — .67ⴱⴱⴱ ⫺.08ⴱ .07
9. T4 Test performance .18ⴱⴱⴱ ⫺.19ⴱⴱⴱ ⫺.33ⴱⴱⴱ .20ⴱⴱⴱ ⫺.21ⴱⴱⴱ ⫺.38ⴱⴱⴱ .20ⴱⴱⴱ .69ⴱⴱⴱ — ⫺.06 .10ⴱⴱ
10. Gender ⫺.16ⴱⴱⴱ .05 .10ⴱⴱ ⫺.16ⴱⴱⴱ .11ⴱⴱⴱ .09ⴱⴱ ⫺.08ⴱⴱ ⫺.08ⴱⴱ ⫺.08ⴱ — —
11. Age ⫺.06ⴱ .10ⴱⴱⴱ .01 ⫺.01 .03 .02 .01 .07ⴱ .11ⴱⴱ — —
Note. Latent bivariate correlations are shown above the diagonal, manifest Pearson’s correlations below the diagonal.
ⴱ
p ⬍ .05. ⴱⴱ p ⬍ .01. ⴱⴱⴱ p ⬍ .001.
mance was regressed onto T1 and T3 classroom emotions, and T2 ance model constrained factor loadings of items to be equal, the
mathematics test performance, and T2 mathematics test perfor- scalar invariance model additionally constrained intercepts of
mance regression into on T1 and T3 classroom emotions, in SPSS. items to be equal, and the strict invariance model additionally
Tolerance statistics were ⬎ .33 and variance inflation fac- constrained item residual variances of items to be equal. Tests of
tors ⬍3.00, suggesting that multicollinearity would not unduly measurement invariance are reported in Table 3. A decline in
bias parameter estimates. model fit of ⌬RMSEA ⬎ .015 or ⌬CFI/TLI ⬎ .01, from one model
to the next, indicates noninvariance (Chen, 2007; Cheung & Rens-
Measurement Invariance vold, 2002). All classroom achievement emotions showed strict
measurement invariance, and so we proceeded to examine longi-
A prerequisite for the modeling of longitudinal relations is tudinal relations.
temporal measurement invariance (Widaman, Ferrer, & Conger,
2010). Accordingly, we tested a series of models for classroom
Structural Equation Modeling of Reciprocal Effects
achievement emotions that included successively strict constraints
(Meredith, 1993). The configural model specified the measure- SEM was used to test the fully forward reciprocal relations model
ment model at T1 and T3 and included correlations between the (RRM) shown in Figure 1. This model was tested competitively (see
corresponding indicators at the two time points. The metric invari- Table 4) against three alternative models: a baseline model where all
Table 3
Tests of Measurement Invariance for Classroom Achievement Emotions
Variable ␹2(df) RMSEA SRMR CFI TLI ⌬RMSEA ⌬CFI ⌬TLI
Enjoyment
T1 3.16(2) .022 .006 1.000 .999
T3 3.47(2) .026 .005 .999 .997
Configural 17.89(15) .011 .012 1.000 .999
Metric invariance 22.14(18) .012 .020 .999 .999 ⫹.001 ⫺.001 ⬍.001
Scalar invariance 44.72(22) .026 .037 .996 .995 ⫹.014 ⫺.003 ⫺.004
Residual invariance 38.56(26) .018 .034 .998 .998 ⫺.008 ⫹.002 ⫹.003
Boredom
T1 1.42(2) .000 .004 1.000 1.001
T3 24.81(2) .100 .020 .981 .944
Configural 67.33(15) .048 .023 .988 .977
Metric invariance 76.91(18) .047 .027 .986 .978 ⫺.001 ⫺.002 ⫺.001
Scalar invariance 85.82(22) .044 .036 .985 .981 ⫺.003 ⫺.001 ⫹.003
Residual invariance 88.38(26) .040 .035 .985 .984 ⫺.004 ⬍.001 ⫹.003
Anxiety
T1 2.18(2) .008 .007 1.000 1.000
T3 8.95(2) .055 .018 .990 .997
Configural 40.78(15) .034 .023 .988 .987
Metric invariance 42.26(18) .030 .024 .989 .983 ⫺.004 ⫹.001 ⫺.004
Scalar invariance 42.87(22) .025 .025 .991 .988 ⫺.005 ⫹.002 ⫹.005
Residual invariance 54.03(26) .027 .030 .987 .986 ⫹.002 ⫺.004 ⫺.002
Note. ␹2 statistic for all models statistically significant at p ⬍ .001. CFI ⫽ confirmatory fit index; RMSEA ⫽ root mean error of approximation; SRMR ⫽
standardized root mean residual; TLI ⫽ Tucker-Lewis index.
Table 4
Comparison of the Reciprocal Relations Model to the Baseline and Unidirectional Relations Models
Model ␹2(df) RMSEA SRMR CFI TLI AIC ⌬AIC TRd(df)

ⴱⴱⴱ
Baseline model 729.48 (315) .033 .082 .972 .967 76777.58 209.96 188.28 (18)ⴱⴱⴱ
Unidirectional model A 609.64 (306)ⴱⴱⴱ .028 .052 .980 .975 76647.67 80.05 79.89 (9)ⴱⴱⴱ
Unidirectional model B 686.71 (312)ⴱⴱⴱ .031 .070 .975 .969 76734.32 166.70 140.07 (15)ⴱⴱⴱ
Reciprocal relations model 526.24 (297)ⴱⴱⴱ .025 .027 .985 .980 76567.62 — —
Note. AIC ⫽ Akaike information criterion; CFI ⫽ confirmatory fit index; RMSEA ⫽ root mean error of approximation; SRMR ⫽ standardized root mean
residual; TLI ⫽ Tucker-Lewis index; TRd ⫽ Satorra–Bentler scaled ␹2 difference. Unidirectional Model A: Relations of emotions to subsequent
performance freely estimated, relations of performance to subsequent emotions constrained to zero. Unidirectional Model B: Relations of performance to
subsequent emotions freely estimated, relations of emotions to subsequent performance constrained to zero.
ⴱⴱⴱ
p ⬍ .001.
directional paths were constrained to zero, a unidirectional model cross-lagged relations with prior T1 boredom and T1 anxiety. T2
where paths from classroom achievement emotions to test perfor- mathematics test performance negatively predicted T3 boredom over
mance were freely estimated but those from test performance to and above the auto-lagged relation with prior T1 boredom and the
achievement emotions were constrained to zero, and a unidirectional cross-lagged relations with prior T1 enjoyment and T1 anxiety. T2
model where paths from test performance to achievement emotions mathematics test performance also negatively predicted T3 anxiety
were freely estimated but those from achievement emotions to test over and above the auto-lagged relation with prior T1 anxiety and the
performance were constrained to zero. All models included age and cross-lagged relations with prior T1 enjoyment, and T1 boredom.
gender as covariates. The RRM showed a good fit to the data that was Relations from T3 emotions to T4 test performance. T3
superior to all other models (see Table 4). Furthermore, the RRM anxiety negatively predicted T4 mathematics test performance over
showed a significantly better fit than the other models using the and above the auto-lagged relation with prior T2 performance. T3
Satorra–Bentler scaled ␹2 difference test (TRd; Bryant & Satorra, enjoyment, T3 boredom, T1 enjoyment, T1 boredom, and T1 anxiety
2012) and an improved relative fit on the Akaike information criterion were not significantly related to T4 test performance.
(AIC). Lower AIC values are indicative of a better model fit, and Relations with covariates. Gender showed a significantly neg-
⌬AIC ⬎10 indicates a substantive change in model fit (Hix-Small, ative relation with T1 enjoyment and significantly positive relations
Duncan, Duncan, & Okut, 2004). The RRM (see Figure 2) was with T1 boredom and anxiety. Age was significantly negatively re-
accepted as the better fitting model. Standardized path coefficients lated to T1 enjoyment and positive related to T1 boredom.
and standard errors are reported in Table 5 and were interpreted as ␤s
.05 to .09 indicating small effects, .10 to .24 moderate effects, Moderating Effects of Academic Buoyancy in the
and ⬎.25 large effects (Keith, 2006).
Relation Between Emotions and Test Performance
Relations from T1 emotions to T2 test performance. T1
anxiety negatively predicted T2 mathematics test performance. T1 enjoy- Interactions for Academic Buoyancy ⫻ Enjoyment, Academic
ment and T1 boredom were not significantly related to T2 performance. Buoyancy ⫻ Boredom, and Academic Buoyancy ⫻ Anxiety, were
Relations from T2 test performance to T3 emotions. T2 estimated in a single model using the latent moderated structural
mathematics test performance positively predicted T3 enjoyment over equation modeling (LMS) approach (Klein & Moosbrugger, 2000)
and above the auto-lagged relation with prior T1 enjoyment and the implemented in Mplus v.8 (Muthén & Muthén, 2017). Absolute
T1 Enjoyment .59 T3 Enjoyment
-.31 .12
-.68
-.68
.60
-.36
T1 Boredom T2 Maths Test -.14 T3 Boredom T4 Maths Test
-.16
-.35
-.18
.64
.60
-.36
.24 -.19
T1 Anxiety .50 T3 Anxiety

-.19
Figure 2. Significant autoregressive, cross-lagged, and concurrent relations between achievement emotions
and test performance. Paths for gender and age are not depicted.
Table 5
Standardized Path Coefficients for the Fully-Forward Reciprocal Relations Model (Standard Errors in Parentheses)
T4 Test
Variable T1 Enjoyment T1 Boredom T1 Anxiety T2 Test performance T3 Enjoyment T3 Boredom T3 Anxiety performance
T1 Enjoyment .09 (.06) .59 (.05) ⫺.31 (.07) ⫺.19 (.08) .05 (.05)
T1 Boredom ⫺.11 (.07) .04 (.06) .24 (.07) ⫺.18 (.07) .08 (.08)
T1 Anxiety ⫺.36 (.06) ⫺.03 (.05) ⫺.05 (.06) .50 (.07) ⫺.13 (.08)
T2 Test performance .12 (.04) ⫺.14 (.03) ⫺.19 (.04) .60 (.03)
T3 Enjoyment ⫺.05 (.06)
T3 Boredom ⫺.01 (.05)
T3 Anxiety ⫺.16 (.07)
Gender ⫺.17 (.03) ⫺.07 (.03) .12 (.04) ⫺.05 (.03) ⫺.03 (.04) .02 (.03) ⫺.01 (.03) .01 (.04)
Age ⫺.07 (.03) .11 (.03) .01 (.04) .06 (.04) .02 (.03) .01 (.03) .03 (.04) .05 (.04)
model fit indices are not provided in the LMS approach. However, demic Buoyancy ⫻ Boredom, and Academic Buoyancy ⫻ Anxi-
it is possible to establish whether a model including an interaction ety) showed an improved fit (⌬AIC ⫽ ⫺4.31, ⌬aBIC ⫽ ⫺3.97)
term offers a better fit to the data using relative fit indices and explained a greater proportion of variance in T4 test perfor-
(Maslowsky, Jager, & Hemken, 2015): Akaike Information Crite- mance (⌬R2 ⫽ .01). Furthermore, a statistically significant likeli-
rion (AIC), sample-size adjusted Bayesian information criterion hood ratio test, D(3) ⫽ 22.26, p ⬍ .001, indicated a worse fit for
(aBIC), the log likelihood ratio test, and the change in the propor- the model without the three interaction terms. Structural coeffi-
tion of variance (⌬R2) explained in the outcome variable (i.e., T4 cients are shown in Table 6. A statistically significant interaction
mathematics test performance). was shown for Academic Buoyancy ⫻ Anxiety, but not Academic
A better model fit is indicated by smaller AIC and aBIC values Buoyancy ⫻ Enjoyment or Academic Buoyancy ⫻ Boredom.
(Hix-Small et al., 2004) and a larger R2 in explaining the variance To probe the Academic Buoyancy ⫻ Anxiety interaction
of the outcome. A statistically significant log likelihood ratio test simple slopes were estimated at ⫾1 SD buoyancy (Figure 3a;
(D) would indicate a worse fit for the more parsimonious model
please note that the Mplus software estimates simple slopes as
(i.e., the model without the interaction term). Because of the
unstandardized regression coefficients only). At ⫹1 SD buoy-
computational power required for the LMS approach 5,000 Monte
ancy a negative relation was shown between anxiety and per-
Carlo Integration points were used.
formance (B ⫽ ⫺3.91, SE ⫽ 1.27). This relation was weaker at
Age and gender were included as covariates. Because we were
mean buoyancy (B ⫽ ⫺1.38, SE ⫽ .34), and at ⫺1 SD buoy-
predicting T4 test performance, autoregressive relations with T2
test performance were also controlled for. Because T3 emotions ancy became nonsignificant (B ⫽ 1.16, SE ⫽ 1.25). To facili-
were predictor variables in these analyses, there would have been tate interpretation of the interaction, we also plotted relations
no benefit to including T1 emotions in the models. Hence, to keep between academic buoyancy and test performance at ⫾1 SD
the models as parsimonious as possible, T1 emotions were not anxiety (see Figure 3b). At ⫹1 SD anxiety, a negative relation
included. In the log likelihood ratio test the three additional pa- was shown between academic buoyancy and test performance
rameters, one per interaction term, equate to three degrees of (B ⫽ ⫺2.80, SE ⫽ 1.38). This relation was not significant at
freedom. mean anxiety (B ⫽ - .01, SE ⫽ .21) and became positive at ⫺1
A model without an interaction term showed a good fit to the SD anxiety (B ⫽ 2.20, SE ⫽ 1.33). The findings suggest that
data: ␹2(146) ⫽ 422.44, p ⬍ .001, RMSEA ⫽ .036, SRMR ⫽ performance benefited from lower anxiety combined with
.031, CFI ⫽ .971, and TLI ⫽ .962. However, the model including higher buoyancy. At higher levels of anxiety the performance-
three interaction terms (Academic Buoyancy ⫻ Enjoyment, Aca- enhancing influence of buoyancy declined.
Table 6
Standardized Path Coefficients for the LI-SEM to Predict T4 Test Performance From Interactions Between Academic Buoyancy and
Enjoyment, Boredom, and Anxiety (Standard Errors in Parentheses)
Predictor T2 Test performance Enjoyment Boredom Anxiety Academic buoyancy T4 Test performance
Enjoyment (ENJ) .04 (.06)

Boredom (BOR) ⫺.08 (.05)
Anxiety (ANX) ⫺.25 (.06)
Academic buoyancy (B) .01 (.04)
B ⫻ ENJ ⫺.05 (.06)
B ⫻ BOR ⫺.03 (.07)
B ⫻ ANX ⫺.10 (.04)
T2 Test performance .19 (.04) ⫺.19 (.04) ⫺.34 (.04) .25 (.05) .61 (.03)
Gender ⫺.06 (.05) .21 (.07) .08 (.09) .06 (.07) .03 (.05) .02 (.06)
Age .06 (.04) .06 (.09) .10 (.08) .02 (.06) .01 (.07) .10 (.07)
Note. LI-SEM ⫽ latent interaction structural equation model.
a 12
10
T4 Mathematics Test Performance

8
6
4 T3 Academic Buoyancy
2 Low (-1SD)
Mean
0
High (+1SD)
-2
-4
-6
-8
-2 SD -1.5 SD -1 SD -0.5 SD Mean +0.5 SD +1 SD +1.5 SD +2 SD
T3 Anxiety
b
T4 Mathematics Test Performance
3 T3 Anxiety
Low (-1SD)
-2 Mean
High (+1SD)
-7
-12
-2 SD -1.5 SD -1 SD -0.5 SD Mean +0.5 SD +1 SD +1.5 SD +2 SD
T3 Academic Buouancy
Figure 3. (a) The model-implied effect of the T3 Academic Buoyancy ⫻ Anxiety interaction on T4 mathe-
matics test performance. Anxiety is represented on the x axis and slopes plotted for ⫾1 SD academic buoyancy.
(b) The model-implied effect of the T3 Academic Buoyancy ⫻ Anxiety interaction on T4 mathematics test
performance. Academic buoyancy is represented on the x axis and slopes plotted for ⫾1 SD anxiety.
Discussion higher anxiety was related to subsequent lower test scores, and
lower test scores, in turn, to higher subsequent anxiety. Overall,
Using a longitudinal design with four alternating waves of data
this pattern of findings suggests that when multiple achievement
collection, we investigated the relations between three classroom
emotions are considered simultaneously, anxiety exerts a stronger
achievement emotions (enjoyment, boredom, and anxiety) and test
performance in the context of mathematics at primary school. effect on performance than enjoyment and boredom. It is notable
Supporting Hypothesis 1, anxiety negatively predicted subsequent that despite low mathematics test scores at T2, there was sufficient
mathematics test scores after controlling for gender, age, concur- variance with which to model relations with previous and subse-
rent classroom achievement emotions, and autoregressive relations quent classroom emotions.
with prior achievement. Enjoyment and boredom, however, were We also investigated whether academic buoyancy moderated
not significantly related to subsequent test scores. In line with relations between negative emotions that are detrimental to learn-
Hypothesis 2, test scores positively predicted subsequent enjoy- ing (boredom and anxiety) and protected test scores from these
ment and negatively predicted boredom and anxiety after control- emotions. The model including interaction terms showed a supe-
ling for prior emotions and the covariates. Thus, a positive feed- rior fit relative to a model not including interactions and a signif-
back loop over time was shown between anxiety and test scores; icant effect was shown for the Academic Buoyancy ⫻ Emotion
interaction. In partial support of Hypothesis 3, test performance trinsic motivation, and depth of learning (Fredrickson, 2001;
was highest when anxiety was low and buoyancy high. With Fredrickson & Branigan, 2005). Practically speaking, when mul-
higher anxiety, the benefit for performance shown for high aca- tiple emotions are considered simultaneously in achievement set-
demic buoyancy diminished. tings, or when the presence of unpleasant emotions is associated
with the absence of pleasant emotions, it may be more important
Relations Between Achievement Emotions and to focus on reducing anxiety in the first instance than to attempt to
reduce boredom and foster enjoyment.
Academic Achievement
In summary, we found support for our second hypothesis, that
CVT proposes reciprocal relations between achievement emo- test performance would be positively related to subsequent enjoy-
tions and indicators of achievement, such as test scores, as mea- ment and negatively related to subsequent boredom and anxiety.
sured in the present study. Emotions exert cognitive-motivational There was, however, only mixed support for our first hypothesis,
effects that influence learning and performance, and learning and that enjoyment would be positively, and boredom and anxiety
performance, in turn, reinforce the control-value appraisals that negatively, related to subsequent test performance. Our findings
underpin emotions (Pekrun, 2006, 2017; Pekrun & Perry, 2014). build on the body of evidence to show that anxiety is detrimental
Previous studies have shown how achievement emotions relate to for mathematics achievement and that it is not just an outcome or
subsequent mathematics achievement in primary/elementary, sec- epiphenomenon of prior performance, but can predict lower
ondary, and undergraduate students (e.g., Ahmed et al., 2013; achievement beyond the influence of prior performance.
Raccanello et al., 2019; Villavicencio, & Bernardo, 2016) but far A minor but nonetheless intriguing point is that T1 enjoyment
fewer have examined how emotions and mathematics achievement was a stronger predictor of T3 boredom than T1 boredom. We
are reciprocally related over time (Pekrun et al., 2014, for an suspect that this finding is attributable to students’ enjoyment
undergraduate psychology course, Pekrun et al., 2014, for second- being more stable over time as compared with boredom (see the
ary school students, and Putwain, Becker, et al., 2018, for primary autoregressive effects for the two emotions from T1 to T3, .59 and
school students). .24, respectively; Table 5). The high stability of enjoyment may
A question of conceptual and applied importance is whether have made it possible for enjoyment to more continuously influ-
achievement emotions exert unique effects when considered si- ence students’ subsequent boredom. The exact reason why enjoy-
multaneously. Notably, only one study thus far has examined ment was more than stable than boredom is unclear but could be
reciprocal relations with mathematics achievement for more than a related to the differential role of perceived control as an antecedent
single emotion in tandem; Putwain, Becker, et al. (2018) examined of the two emotions. Perceived control will result in greater en-
enjoyment and boredom simultaneously and found that unique joyment and will also lead to lower boredom in most cases, but can
reciprocal effects for both of these emotions. The findings of the alternatively induce higher boredom owing to a lack of challenge
present study offer novel insights into the question of unique (e.g., Pekrun, 2006, 2018; Pekrun & Perry, 2014), thus explaining
reciprocal effects by considering three classroom emotions (enjoy- lower stability of boredom.
ment, boredom, and anxiety) simultaneously in a single analytic It could be questioned whether findings from the present study
model. When the variance shared between emotions measured would generalize to nontesting situations typically encountered by
concurrently was controlled for in this way, reciprocal relations students during classroom learning. There is evidence that domain-
with mathematics achievement were only shown for anxiety. Al- specific emotions related to learning (e.g., in math, science, and L2
though mathematics test performance predicted subsequent enjoy- learning) can influence classroom achievement (for reviews see
ment and boredom (in the expected direction), enjoyment and Horwitz, 2001; Maloney, 2016; Sinatra, Broughton, & Lombardi,
boredom were not significantly related to subsequent test perfor- 2016). For emotions related to classroom activities specifically,
mance. class-related anxiety has been shown to relate negatively to both
Thus, when these three emotions are pitted competitively class grades and test scores in mathematics (rs ⫽ ⫺.21 to ⫺.32;
against each other, anxiety exerts the strongest predictive power Lichtenfeld et al., 2012; Peixoto, Sanches, Mata, & Monteiro,
for achievement. It is important to clarify that we do not argue 2017; Putwain et al., 2020; Raccanello et al., 2019; for L2 class-
against the presence of reciprocal relations between enjoyment and related anxiety, see Shao, Pekrun, Marsh, & Loderer, 2020). In the
achievement, and between boredom and achievement. Rather, the study by Lichtenfeld et al. (2012), the magnitude of correlations
findings suggest that anxiety emerges as the most significant between classroom-related anxiety and two different measures of
(statistically and practically) of the three emotions when they are math achievement (class grades vs. test scores) did not differ
modeled together. Analytically speaking, this is a combined result (z ⫽ ⫺.40 and .43 for 2nd and 3rd grade samples, respectively;
of anxiety being related more strongly to test performance than ps ⬎.05). Overall, these findings would suggest the results of the
enjoyment or boredom and of the intercorrelations between enjoy- present study for relations between class-related emotions, notably
ment, boredom, and anxiety. The result is that insufficient unique anxiety, and test scores would be applicable to classroom achieve-
variance remained in enjoyment and boredom to predict test per- ment, such as grades based on classroom activities, as well.
formance.
Theoretically speaking, the implication is that the cognitive-
The Moderating Effect of Academic Buoyancy
motivational mechanisms affected by anxiety, specifically inter-
ference with working memory capacity and function (Eysenck & Academic buoyancy is the capacity to respond effectively to
Derakshan, 2011; Eysenck et al., 2007), may exert a stronger minor academic adversities (Martin & Marsh, 2009). Studies have
influence on performance than the cognitive-motivational mecha- shown that higher domain general academic buoyancy is related to
nisms affected by enjoyment and boredom, such as interest, in- adaptive educational outcomes in secondary school students (e.g.,
Malmberg, Hall, & Martin, 2013; Martin et al., 2013) including of test-related emotions occurring during test taking. Furthermore,
higher positive, and lower negative, achievement emotions in measures of engagement and learning during mathematics lessons
primary and secondary school students (e.g., Hirvonen et al., 2019; (e.g., on-task behaviors, cognitive strategies used, and tasks com-
Martin et al., 2010; Putwain et al., 2012). Studies linking domain pleted) may have been more sensitive to the protection offered by
general academic buoyancy to aggregated achievement in mathe- buoyancy against classroom learning anxiety, than test perfor-
matics and reading have shown equivocal results, however, with mance.
some reporting statistically significant relations but others not Third, it is known that children can overestimate their abilities
(e.g., Collie et al., 2015; Martin, 2014). We reasoned this may be (see Muenks, Wigfield, & Eccles, 2018; Salles, Ais, Semelman,
partly an artifact of reduced predictive power arising from lack of Sigman, & Calero, 2016). It is possible that participants may have
domain specificity and used a mathematics-specific measure of overestimated their academic buoyancy and believed they had a
academic buoyancy in the present study (also ensuring matching greater capacity to bounce back than was actually the case. This
specificity with domain-specific achievement emotions). may account for the diminishing protection offered by academic
Conceptually, higher levels of buoyancy could moderate (i.e., buoyancy as anxiety increased; there was less buoyancy than was
buffer) the impact of academic adversities on subsequent outcomes reported. This overestimation could also account for why very high
(e.g., Martin & Marsh, 2019; Putwain et al., 2016; Symes et al., (⫹2 SD) academic buoyancy became detrimental to performance.
2015). Accordingly, we theorized that higher levels of buoyancy If participants anticipated they would be highly able to effectively
could protect test performance from boredom and anxiety. The respond to difficulties encountered during the test but experienced
model including the interaction terms between academic buoyancy the opposite, they could be highly demotivated, reduce effort, or
and emotions showed a superior fit to the model not including even give up completely. This is somewhat akin to choking when
these interactions and a statistically significant interaction between the pressure from performing mathematics tasks becomes exag-
Academic Buoyancy and Anxiety. Interactions between Academic gerated by concerns about one’s performance overloading working
Buoyancy and Enjoyment, and between Academic Buoyancy and memory (e.g., Beilock, Kulp, Holt, & Carr, 2004).
Boredom, were not statistically significant. Fourth, a curious finding reported by Martin and Marsh (2019)
The pattern of the observed interaction partially supported our was that academic buoyancy exerted a greater protective role for
third hypothesis. Rather than protecting test performance at higher academic adversities reported 12 months later than with current
levels of anxiety, higher academic buoyancy amplified test perfor- academic adversities. This finding is in line with the view that
mance at lower levels of anxiety. With higher anxiety, the differ- some exposure to adversity is necessary for persons to build
ential benefit to performance offered by higher academic buoy- adaptive responses. It is therefore possible that the benefits of
ancy diminished. This finding suggests that academic buoyancy buoyancy play out, or accumulate, over time rather than contem-
did not protect test performance against the adverse effects of high poraneously. If this were the case, then we would have been
classroom anxiety. Furthermore, there is the intriguing finding that unlikely to observe a protective effect on a test taken one week
high buoyancy may even be detrimental for performance when after measurements of self-reported anxiety and academic buoy-
anxiety is also high. We offer four considerations to explain this ancy. In this respect academic buoyancy can be considered as a
pattern of effects. malleable trait, a lasting attribute that can develop over time and
First, buoyancy may not have protected performance because that is responsive to intervention (Martin, 2013b). Beneficial ef-
anxiety was too adverse. Although academic anxiety is often fects would occur downstream at a later point in time. We con-
considered a minor form of adversity in comparison to clinical clude the theoretical proposition that academic buoyancy can
anxiety (Martin, 2013a), it is known that high levels of academic protect subsequent outcomes from minor adversity may be valid
anxiety overlap with clinical anxiety (e.g., Herzer, Wendt, & despite having been confirmed only partially in the present study.
Hamm, 2014; Warren, Ollendick, & King, 1996; see also Pekrun Although we focused on mathematics in the present study,
& Loderer, in press). It may be that the higher levels of anxiety because we wanted to adopt a domain-specific approach, other
measured in the present study constituted more of a major than a studies on buoyancy have considered English and reading in
minor adversity to students, in which case buoyancy may not have research with primary and secondary students (e.g., Colmar et al.,
been a sufficiently strong factor to offer protection. Classroom 2019; Putwain & Aveyard, 2018). There are no theoretical reasons
anxieties may have been furthered by the subject material; math- for academic buoyancy to differentially relate to achievement and
ematics learning and reasoning is, for many students, a source of adaptive beliefs, emotions, and behaviors, in varying academic
anxiety (e.g., Maloney, 2016). Martin (2013a) and Martin and subjects. Although mathematics may be anxiety-provoking for
Marsh (2008, 2009) argue that resilience, rather than buoyancy, is some students, it is known that literacy also presents a challenge
required to respond to chronic or stronger academic adversities. for students, and reading motivation begins to decline at the end of
Second, an alternative interpretation would be that buoyancy did elementary school (see Wigfield, 1997). These are adversities that
not protect against performance because the classroom measure of buoyancy is theorized to protect against. However, we are mindful
anxiety underestimated the level of anxiety experienced during the that some subjects do present unique challenges for students and it
test. Although the mathematics test used in the present study was would be beneficial for academic buoyancy studies to broaden the
characterized as low-stakes, it is possible that a greater degree of repertoire of academic subjects considered.
anxiety was experienced during the test than was typically expe- Despite a positive bivariate correlation between academic buoy-
rienced during classroom learning. In this respect the classroom ancy and subsequent test performance, there was no first-order
measure may have reported anxiety as not being sufficiently ad- effect of academic buoyancy on performance after controlling for
verse for academic buoyancy to effectively offer protection. To current relations with emotion, the interaction with emotion, and
investigate this possibility, future research could include a measure prior achievement. The prediction made in Hypothesis 1 that
buoyancy would be positively related to achievement was not studies may follow a similar approach to test and report assump-
supported in these models. It is possible that emotions mediate tions of MCAR and MAR more completely and openly.
relations between academic buoyancy and subsequent achieve-
ment (see Putwain et al., 2015). Although we could not formally Limitations and Directions for Future Research
test this hypothesis because academic buoyancy and achievement
emotions were assessed simultaneously at T3, this interpretation is Despite the novel theoretical contributions of this study and the
supported by the strong relations between buoyancy and emotions use of a relatively understudied age group in achievement emotion
(see Table 1), combined with the relations between emotions and and academic buoyancy research, there are limitations that should
subsequent test performance. Hence, the relations with emotions be considered and can be used to suggest directions for subsequent
may have reduced the direct predictive power of academic buoy- studies. First, we were able to measure data across two alternate
ancy. In this case, academic buoyancy would represent a relatively assessments of emotions and performance. Thus, we were able to
rare example of the same variable operating as both a mediator and test the predictive power of emotions on subsequent performance
a moderator (for a related example for self-efficacy as a mediator twice but reciprocal predictive effects of test performance on
and moderator, see Dicke et al., 2014). subsequent emotions only once. Although this design is sufficient
When considering findings, it is important to acknowledge that to test reciprocal relations (Little et al., 2007; Rosel & Plewis,
the mathematics tests used in the present study contained only 2008), additional alternating waves of emotions and achievement
closed response questions that required convergent styles of rea- would allow for multiple assessments of reciprocal relations be-
soning. Some students may have employed the correct approach to tween emotions and achievement. The present design only permit-
solving a question but ultimately arrived at an incorrect answer. ted a single test for the moderating effect of academic buoyancy
Awarding marks for reasoning (as well as for proving a correct while controlling for previous achievement. Additional alternating
answer) or including open questions in addition to, or instead of, waves of academic buoyancy (alongside emotions) and achieve-
closed questions could potentially show a different pattern of ment would also allow for multiple tests of protective effects of
relations with classroom emotions and academic buoyancy. Sim- buoyancy on relations between adverse emotions and achievement.
ilarly, questions requiring creative and global processing that re- A related point is that there were unequal time intervals between
assessments. T1 and T2, and T3 and T4, were spaced apart by one
quires making new connections between concepts, which are po-
week. T2 and T3, however, were spaced apart by approximately
tentially more challenging, could also relate differently to
seven months. We were constrained by school administration to
classroom emotions and academic buoyancy than questions requir-
schedule data collection this way in order to minimize impact on
ing convergent analytical reasoning.
routine teaching and learning. Thus, we cannot make direct com-
We would also like to briefly comment on the missing data in
parisons of the size of paths from emotions to test performance
our study and how it was treated. Analyses suggested that the
(with the one week time interval) with the size of the paths from
cause of the missing data were T2 mathematics test scores (and
test performance to emotions (with the seven month time interval).
hence missing data were treated as MAR). The implication of
Second, we measured three classroom emotions (enjoyment,
MAR is that missingness was not completely random, but could be
boredom, and anxiety). Although being the first study to model
treated as random after T2 mathematics scores were controlled for
reciprocal relations in all three simultaneously, there are other
(see Little & Rubin, 2002). This situation is typical of applied achievement emotions likely to co-occur in classroom settings
research where substantive study variables or sociodemographic (e.g., hope and hopelessness). It would be useful to include addi-
correlates may influence decisions whether to continue participa- tional emotions to address the question of which emotions are the
tion or not (e.g., Lamers et al., 2012). Attrition may be higher for most meaningful predictors of achievement, when considered to-
groups of participants characterized by vulnerability, low motiva- gether. We were somewhat limited in that the only validated
tion, or where the study may pose emotional distress. In the present measure of achievement emotions available for the age group of
study, it would seem plausible that students who experienced the participants used in the present study (AEQ-ES; Lichtenfeld et al.,
T2 mathematics test as more difficult (and hence performed worse) 2012) measures enjoyment, boredom, and anxiety specifically. It
may have been less motivated to continue participating because would be a useful extension of this measure to include additional
this may have reminded them of their difficulties with mathemat- emotions.
ics. Third, from a CVT perspective, emotions influence subsequent
When the cause of the missing data is ignored, there is a danger achievement through cognitive-motivational mechanisms and
that model estimates may be biased. In the present study, the risk achievement influences subsequent emotions through control-
would be an underrepresentation of participants with lower math- value appraisals. Although it is useful to first test for reciprocal
ematics scores. However, simulation studies have shown that when relations solely between emotions and achievement, partly because
the cause of missing data is included in the algorithm used for these relations are meaningful in their own right and partly as a
handling missing data (FIML was used in the present study), precursor to investigating mediating processes, future studies
disproportional participant attrition can be corrected for to yield should additionally include cognitive-motivational mechanisms
unbiased estimates (e.g., Collins, Schafer, & Kam, 2001; Nichol- and control-value appraisals. A study that combined tests of re-
son et al., 2017). We followed a strategy to identify the cause of ciprocal effects in conjunction with the presumed mediating mech-
the missing data (T2 mathematics scores) and included that vari- anisms would offer an even more robust test of CVT.
able in our analytic models. Hence, we are confident that param- Fourth, we speculated two reasons for the inconsistent relations
eter estimates are applicable to the entire sample despite reduced shown been academic buoyancy and achievement in the extant
representation of participants with lower T2 test scores. Future literature; low domain specificity or mismatch between measures
of academic buoyancy and achievement, and the presence of cally interventions have focused on test anxiety (von der Embse,
additional variables that may either overlap with academic buoy- Barterian, & Segool, 2013), school phobia, and school refusal
ancy (e.g., grit or future time perspective; Fong & Kim, 2019) or (Lauchlan, 2003). There are fewer interventions for classroom or
mediate relations between academic buoyancy and achievement learning anxiety, and these are focused on specific forms of anx-
(e.g., perceived control; Collie et al., 2015, and Putwain & Ave- iety such as math anxiety (Schaeffer, Rozek, Berkowitz, Levine, &
yard, 2018). The correlations between domain-specific academic Beilock, 2018) and statistics anxiety (Smith & Capuzzi, 2019).
buoyancy in mathematics and students’ mathematics achievement Math anxiety interventions are germane to the present study with
found in the present study (rs ⫽ .23 and .25) were stronger than the substantive focus on classroom emotions in mathematics (cog-
correlations between domain-general academic buoyancy and nate, although not identical with math anxiety).
achievement in previous studies (e.g., rs ⫽ .13–.17; Martin, 2014; Math anxiety interventions have focused broadly on either
Putwain et al., 2015). This would lend credibility to the view that building subject mastery, reducing negative appraisals of compe-
high domain specificity between academic buoyancy and achieve- tence or physiological arousal, or the normalization of failure as
ment strengthens relations. Furthermore, in the LI-SEM, the first- part of learning (Ramirez, Shaw, & Maloney, 2018). In CVT
order effects of academic buoyancy were negligible; this indicates (Pekrun, 2016, 2017; Pekrun & Perry, 2014) and the integrated
a possible mediating role of emotions. However, from the present model of emotion regulation in achievement situations proposed
study we cannot establish whether domain-general variables that by Harley, Pekrun, Taxer, and Gross (2019), mastery, reappraisal,
overlap with academic buoyancy, such as grit and future time and failure-response beliefs correspond to ways of building per-
perspective, would reduce the magnitude of relations between ceived control or regulating emotion through cognitive change.
academic buoyancy and achievement when using domain-specific These theoretically derived mechanisms of reducing anxiety are
measures of academic buoyancy. Future studies should continue to not solely the province of psychologists and specialist interven-
explore how academic buoyancy (especially when considered as a tions. There are practical ways in which instructors can incorporate
domain-specific construct) can be differentiated from cognate con- mastery practice and adaptive responses to failure in both mathe-
structs and how relations with achievement can be used to inform matics and other subjects through directing student attributions for
understanding of overlap between similar constructs. success and failure (Perry et al., 2014) and creating a classroom
Fifth, we did not account for learning disabilities (e.g., dys- culture whereby failure is defined as a normal part of the learning
graphia or dyscalculia) in the present study. It is likely that they process (Murphy & Dweck, 2010).
would have been meaningful covariates and might explain vari-
ance in both predictors and outcomes. To collect accurate data for
learning disabilities for children aged 9 to 10 years, it would be Conclusion
necessary to use official school records rather than to rely on
When the relations between three classroom emotions (enjoy-
participant self-report; not all students at this age may understand
ment, boredom, and anxiety) and test performance were modeled
if they have been diagnosed with a learning disability or if they
simultaneously over four waves of data collection, reciprocal re-
have, what that learning disability is. To keep data collection
lations were shown for anxiety (in terms of positive reciprocal
anonymous for ethical reasons, we were unable to match partici-
relations). Higher test performance predicted higher enjoyment
pant self-report data with school records. However, it would be
and lower boredom, but enjoyment and boredom were not signif-
desirable for future studies, where ethical protocols permit, to
icantly related to subsequent test performance. Thus, when the
include information about learning disabilities.
shared variance between these emotions is considered (as is likely
Finally, our test of the moderating effect of buoyancy on the
relations between academic adversity and subsequent outcomes to happen in classroom situations where discrete emotions may
was limited to test performance. There are other salient outcomes co-occur), anxiety emerged as the emotion that is more important
that academic buoyancy could protect from adversity, including for achievement. We also investigated whether academic buoy-
attendance, compliance with school behavioral policy, positive ancy might protect performance against anxiety. We found that
relationships with peers and staff, adaptive motivation (e.g., in- this was partially the case. Buoyancy protected performance at
trinsic motivation), and behavioral engagement (e.g., participation lower levels of anxiety, suggesting that buoyancy can help students
in lessons and extracurricular activities). Future research should coping at least with mild forms of negative emotion.
consider such outcomes, and where possible use official school
recorded data (e.g., attendance) to complement self-report data on References
these outcomes. It would also be useful to include measures of
emotional self-regulation to establish whether the moderating in- Ahmed, W., van der Werf, G., Kuyper, H., & Minnaert, A. (2013).
fluence of academic buoyancy is related to differential use of Emotions, self-regulated learning, and achievement in mathematics: A
regulatory strategies. growth curve analysis. Journal of Educational Psychology, 105, 150 –
161. http://dx.doi.org/10.1037/a0030160
Beilock, S. L., Kulp, C. A., Holt, L. E., & Carr, T. H. (2004). More on the
Insights for Practice fragility of performance: Choking under pressure in mathematical prob-
lem solving. Journal of Experimental Psychology: General, 133, 584 –
In classroom settings, when emotions such as enjoyment, bore- 600. http://dx.doi.org/10.1037/0096-3445.133.4.584
dom, and anxiety may co-occur, anxiety emerges as the strongest Bohn-Gettler, C. M. (2019). Getting a grip: The PET framework for
(negative) emotional predictor of achievement according to the studying how reader emotions influence comprehension. Discourse Pro-
present findings. Attempts to reduce anxiety would therefore be cesses, 56, 386 – 401. http://dx.doi.org/10.1080/0163853X.2019
beneficial in helping students’ learning and performance. Typi- .1611174
Brooks, J. E. (2006). Strengthening resilience in children and youths: Fong, C. J., & Kim, Y. W. (2019). A clash of constructs? Re-examining
Maximizing opportunities through the schools. Children & Schools, 28, grit in light of academic buoyancy and future time perspective. Current
69 –76. http://dx.doi.org/10.1093/cs/28.2.69 Psychology. Advance online publication. http://dx.doi.org/10.1007/
Bryant, F. B., & Satorra, A. (2012). Principles and practice of scaled s12144-018-0120-4
difference chi-square testing. Structural Equation Modeling, 19, 372– Forgas, J. P. (2008). Affect and cognition. Perspectives on Psychological
398. http://dx.doi.org/10.1080/10705511.2012.687671 Science, 3, 94 –101. http://dx.doi.org/10.1111/j.1745-6916.2008.00067.x
Chen, F. F. (2007). Sensitivity of goodness of fit indices to lack of Fredrickson, B. L. (2001). The role of positive emotions in positive
measurement invariance. Structural Equation Modeling, 14, 464 –504. psychology: The broaden-and-build theory of positive emotions. Amer-
http://dx.doi.org/10.1080/10705510701301834 ican Psychologist, 56, 218 –226. http://dx.doi.org/10.1037/0003-066X
Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit .56.3.218
indexes for testing measurement invariance. Structural Equation Mod- Fredrickson, B. L., & Branigan, C. (2005). Positive emotions broaden the
eling, 9, 233–255. http://dx.doi.org/10.1207/S15328007SEM0902_5 scope of attention and thought-action repertoires. Cognition and Emo-
Collie, R. J., Martin, A. J., Malmberg, L.-E., Hall, J., & Ginns, P. (2015). tion, 19, 313–332. http://dx.doi.org/10.1080/02699930441000238
Academic buoyancy, student’s achievement, and the linking role of Gable, P., & Harmon-Jones, E. (2010). The motivational dimensional
control: A cross-lagged analysis of high school students. British Journal model of affect: Implications for breadth of attention, memory, and
of Educational Psychology, 85, 113–130. http://dx.doi.org/10.1111/bjep cognitive categorisation. Cognition and Emotion, 24, 322–337. http://dx
.12066 .doi.org/10.1080/02699930903378305
Collins, L. M., Schafer, J. L., & Kam, C. M. (2001). A comparison of Goetz, T., & Hall, N. (2014). Academic boredom. In R. Pekrun & L.
inclusive and restrictive strategies in modern missing data procedures. Linnenbrink-Garcia (Eds.), International handbook of emotions in edu-
Psychological Methods, 6, 330 –351. http://dx.doi.org/10.1037/1082- cation (pp. 311–330). London, UK: Routledge.
989X.6.4.330 Hanushek, E. A., & Woessmann, L. (2008). The role of cognitive skills in
Colmar, S., Liem, G. A. D., Connor, J., & Martin, A. M. (2019). Exploring economic development. Journal of Economic Literature, 46, 607– 668.
the relationships between academic buoyancy, academic self-concept, http://dx.doi.org/10.1257/jel.46.3.607
and academic performance: A study of mathematics and reading among Harley, J. M., Pekrun, R., Taxer, J. L., & Gross, J. J. (2019). Emotion
regulation in achievement situations: An integrated model. Educational
primary school students. Educational Psychology, 39, 1068 –1089.
Psychologist, 54, 106 –126. http://dx.doi.org/10.1080/00461520.2019
http://dx.doi.org/10.1080/01443410.2019.1617409
.1587297
Compas, B. E. (2004). Processes of risk and resilience during adolescence:
Hattie, J., & Timperley, H. (2007). The power of feedback. Review of
Linking contexts and individuals. In R. M. Lerner & L. Steinberg (Eds.),
Educational Research, 77, 81–112. http://dx.doi.org/10.3102/
Handbook of adolescent psychology (pp. 263–296). Hoboken, NJ: Wi-
003465430298487
ley. http://dx.doi.org/10.1002/9780471726746.ch9
Heene, M., Hilbert, S., Draxler, C., Ziegler, M., & Bühner, M. (2011).
Department for Education. (2013). Mathematics programmes of study: Key
Masking misfit in confirmatory factor analysis by increasing unique
Stages 1 and 2 National Curriculum in England. Richmond, UK: Her
variances: A cautionary note on the usefulness of cutoff values of fit
Majesty’s Stationery Office.
indices. Psychological Methods, 16, 319 –336. http://dx.doi.org/10.1037/
Department for Education. (2017). Opportunity areas selection methodol-
a0024917
ogy. Richmond, UK: Her Majesty’s Stationery Office.
Herzer, F., Wendt, J., & Hamm, A. O. (2014). Discriminating clinical from
Department for Education. (2019). National curriculum assessments at Key
nonclinical manifestations of test anxiety: A validation study. Behavior
Stage 2 in England, 2019. Richmond, UK: Her Majesty’s Stationery
Therapy, 45, 222–231. http://dx.doi.org/10.1016/j.beth.2013.11.001
Office.
Hirvonen, R., Putwain, D. W., Määttä, S., Ahonen, T., & Kiuru, N. (2019).
Dicke, T., Parker, P. D., Marsh, H. W., Kunter, M., Schmeck, A., & The role of academic buoyancy and emotions in students’ learning-
Leutner, D. (2014). Self-efficacy in classroom management, classroom related expectations and behaviours in primary school. British Journal of
disturbances, and emotional exhaustion: A moderated mediation analy- Educational Psychology. Advance online publication. http://dx.doi.org/
sis of teacher candidates. Journal of Educational Psychology, 106, 10.1111/bjep.12336
569 –583. http://dx.doi.org/10.1037/a0035504 Hirvonen, R., Yli-Kivistö, L., Putwain, D. W., Ahonen, T., & Kiuru, N.
Di Leo, I., Muis, K. R., Singh, C., & Psaradellis, C. (2019). Curiosity (2019). School-related stress among sixth-grade students–Associations
Confusion? Frustration! The role and sequencing of emotions during with academic buoyancy and temperament. Learning and Individual
mathematics problem solving. Contemporary Educational Psychology, Differences, 70, 100 –108. http://dx.doi.org/10.1016/j.lindif.2019.01.012
58, 121–137. http://dx.doi.org/10.1016/j.cedpsych.2019.03.001 Hix-Small, H., Duncan, T. E., Duncan, S. C., & Okut, H. (2004). A
Duckworth, A. L., Peterson, C., Matthews, M. D., & Kelly, D. R. (2007). multivariate associative finite growth mixture modeling approach exam-
Grit: Perseverance and passion for long-term goals. Journal of Person- ining adolescent alcohol and marijuana use. Journal of Psychopathology
ality and Social Psychology, 92, 1087–1101. http://dx.doi.org/10.1037/ and Behavioral Assessment, 26, 255–270. http://dx.doi.org/10.1023/B:
0022-3514.92.6.1087 JOBA.0000045341.56296.fa
Enders, C. K. (2010). Applied missing data analysis. New York, NY: Holmes, W., & Dowker, A. (2013). Catch up numeracy: A targeted
Guilford Press. intervention for children who are low-attaining in mathematics. Re-
Eysenck, M. W., & Derakshan, N. (2011). New perspectives in attentional search in Mathematics Education, 15, 249 –265. http://dx.doi.org/10
control theory. Personality and Individual Differences, 50, 955–960. .1080/14794802.2013.803779
http://dx.doi.org/10.1016/j.paid.2010.08.019 Horwitz, E. K. (2001). Language anxiety and achievement. Annual Review
Eysenck, M. W., Derakshan, N., Santos, R., & Calvo, M. G. (2007). of Applied Linguistics, 21, 112–126. http://dx.doi.org/10.1017/
Anxiety and cognitive performance: Attentional control theory. Emo- S0267190501000071
tion, 7, 336 –353. http://dx.doi.org/10.1037/1528-3542.7.2.336 Hoy, W. K., Tarter, C. J., & Hoy, A. W. (2006). Academic optimism of schools:
Fiedler, K., & Beier, S. (2014). Affect and cognitive processes in educa- A force for student achievement. American Educational Research Journal, 43,
tional contexts. In R. Pekrun & L. Linnenbrink-Garcia (Eds.), Interna- 425–446. http://dx.doi.org/10.3102/00028312043003425
tional handbook of emotions in education (pp. 36 –55). London, UK: Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance
Taylor & Francis. structure analysis: Conventional criteria versus new alternatives. Struc-
tural Equation Modeling, 6, 1–55. http://dx.doi.org/10.1080/ Martin, A. M. (2013a). Academic buoyancy and academic resilience:
10705519909540118 Exploring ‘everyday’ and ‘classic’ resilience in the face of academic
Jeličič, H., Phelps, E., & Lerner, R. M. (2009). Use of missing data adversity. School Psychology International, 34, 488 –500. http://dx.doi
methods in longitudinal studies: The persistence of bad practices in .org/10.1177/0143034312472759
developmental psychology. Developmental Psychology, 45, 1195–1199. Martin, A. J. (2013b). The Personal Proficiency Network: Key self-system
http://dx.doi.org/10.1037/a0015665 factors and processes to optimize academic development. In D. M.
Kärkkäinen, K., & Vincent-Lancrin, S. (2013). Sparking innovation in McInerney, H. W. Marsh, R. G. Craven, & F. Guay (Eds.), Theory
STEM education with technology and collaboration: A case study of the driving research: New wave perspectives on self-processes and human
HP Catalyst Initiative. OECD Education Working Papers, No. 91. development (pp. 251–286). Charlotte, NC: Information Age Publishing.
OECD Publishing. http://dx.doi.org/10.1787/5k480sj9k442-en Martin, A. J. (2014). Academic buoyancy and academic outcomes: To-
Keith, T. Z. (2006). Multiple regression and beyond. Upper Saddle River, wards a further understanding of students with attention-deficit/
NJ: Pearson Education. hyperactivity disorder (ADHD), students without ADHD, and academic
Klein, A., & Moosbrugger, H. (2000). Maximum likelihood estimation of buoyancy itself. British Journal of Educational Psychology, 84, 86 –107.
latent interaction effects with the LMS method. Psychometrika, 65, http://dx.doi.org/10.1111/bjep.12007
457– 474. http://dx.doi.org/10.1007/BF02296338 Martin, A. J., Colmar, S. H., Davey, L. A., & Marsh, H. W. (2010).
Lamers, F., Hoogendoorn, A. W., Smit, J. H., van Dyck, R., Zitman, F. G., Longitudinal modelling of academic buoyancy and motivation: Do the
Nolen, W. A., & Penninx, B. W. (2012). Sociodemographic and psy- ‘5Cs’ hold up over time? British Journal of Educational Psychology, 80,
chiatric determinants of attrition in the Netherlands Study of Depression 473– 496. http://dx.doi.org/10.1348/000709910X486376
and Anxiety (NESDA). Comprehensive Psychiatry, 53, 63–70. http://dx Martin, A. J., Ginns, P., Brackett, M. A., Malmberg, L.-E., & Hall, J.
.doi.org/10.1016/j.comppsych.2011.01.011 (2013). Academic buoyancy and psychological risk: Exploring recipro-
Lance, C. E., Butts, M. M., & Michels, L. C. (2006). The sources of four cal relationships. Learning and Individual Differences, 27, 128 –133.
commonly reported cutoff criteria: What did they really say? Organiza- http://dx.doi.org/10.1016/j.lindif.2013.06.006
tional Research Methods, 9, 202–220. http://dx.doi.org/10.1177/ Martin, A. J., & Marsh, H. W. (2008). Academic buoyancy: Towards an
1094428105284919 understanding of students’ everyday academic resilience. Journal of
Lauchlan, F. (2003). Responding to chronic non-attendance: A review of
School Psychology, 46, 53– 83. http://dx.doi.org/10.1016/j.jsp.2007.01
intervention approaches. Educational Psychology in Practice, 19, 133–
.002
146. http://dx.doi.org/10.1080/02667360303236
Martin, A. J., & Marsh, H. W. (2009). Academic resilience and academic
Lens, W., & Seginer, S. (2015). Future time perspective and motivation.
buoyancy: Multidimensional and hierarchical conceptual framing of
International Encyclopedia of the Social & Behavioral Sciences, 2,
causes, correlates and cognate constructs. Oxford Review of Education,
561–566. http://dx.doi.org/10.1016/B978-0-08-097086-8.24098-1
35, 353–370. http://dx.doi.org/10.1080/03054980902934639
Lichtenfeld, S., Pekrun, R., Stupnisky, R. H., Reiss, K., & Murayama, K.
Martin, A. M., & Marsh, H. W. (2019). Investigating the reciprocal
(2012). Measuring students’ emotions in the early years: The Achieve-
relations between academic buoyancy and academic adversity: Evi-
ment Emotions Questionnaire-Elementary School (AEQ-ES). Learning
dence for the protective role of academic buoyancy in reducing
and Individual Differences, 22, 190 –201. http://dx.doi.org/10.1016/j
academic adversity over time. International Journal of Behavioral
.lindif.2011.04.009
Development. Advance online publication. http://dx.doi.org/10.1177/
Linnenbrink, E. A. (2006). Emotion research in education: Theoretical and
0165025419885027
methodological perspectives on the integration of affect, motivation and
Maslowsky, J., Jager, J., & Hemken, D. (2015). Estimating and interpreting
cognition. Educational Psychology Review, 18, 307–314. http://dx.doi
latent variable interactions: A tutorial for applying the latent moderated
.org/10.1007/s10648-006-9028-x
Little, R. J. A. (1988). A test of missing completely at random for structural equations method. International Journal of Behavioral Devel-
multivariate data with missing values. Journal of the American Statis- opment, 39, 87–96. http://dx.doi.org/10.1177/0165025414552301
tical Association, 83, 1198 –1202. http://dx.doi.org/10.1080/01621459 Meredith, W. (1993). Measurement invariance, factor analysis and factorial
.1988.10478722 invariance. Psychometrika, 58, 525–543. http://dx.doi.org/10.1007/
Little, R. J., & Rubin, D. B. (2002). Statistical analysis with missing data BF02294825
(2nd ed.). Hoboken, NJ: Wiley-Interscience. http://dx.doi.org/10.1002/ Muenks, K., Wigfield, A., & Eccles, J. S. (2018). I can do this! The
9781119013563 development and calibration of children’s expectations for success and
Little, T. D., Preacher, K. J., Selig, J. P., & Card, N. A. (2007). New competence beliefs. Developmental Review, 48, 24 –39. http://dx.doi
developments in latent variable panel analyses of longitudinal data. .org/10.1016/j.dr.2018.04.001
International Journal of Behavioral Development, 31, 357–365. http:// Muis, K. R., Psaradellis, C., Lajoie, S. P., Di Leo, I., & Chevrier, M.
dx.doi.org/10.1177/0165025407077757 (2015). The role of epistemic emotions in mathematics problem solving.
Loderer, K., Pekrun, R., & Lester, J. (2018). Beyond cold technology: A Contemporary Educational Psychology, 42, 172–185. http://dx.doi.org/
systematic review and meta-analysis on emotions in technology-based 10.1016/j.cedpsych.2015.06.003
learning environments. Learning and Instruction. Advance online pub- Murphy, M. C., & Dweck, C. S. (2010). A culture of genius: How an
lication. http://dx.doi.org/10.1016/j.learninstruc.2018.08.002 organization’s lay theory shapes people’s cognition, affect, and behav-
Malmberg, L.-E., Hall, J., & Martin, A. J. (2013). Academic buoyancy in ior. Personality and Social Psychology Bulletin, 36, 283–296. http://dx
secondary school: Exploring patterns of convergence in English, math- .doi.org/10.1177/0146167209347380
ematics, science, and physical education. Learning and Individual Dif- Muthén, L. K., & Muthén, B. O. (2017). Mplus user’s guide (8th ed.).
ferences, 23, 262–266. http://dx.doi.org/10.1016/j.lindif.2012.07.014 London, UK: Author.
Maloney, E. A. (2016). Math anxiety: Causes, consequences, and reme- National Center for Educational Statistics. (2019). 2019 National assessment
diation. In K. R. Wentzel & D. B. Miele (Eds.), Handbook of motivation of educational progress: Highlighted results at grades 4 and 8 for the
at school (pp. 408 – 424). Oxon, UK: Routledge. nation, states, and districts. U.S. Department for Education. Retrieved from
Marsh, H. W. (2007). Self-Concept theory, measurement and research into https://www.nationsreportcard.gov/highlights/mathematics/2019
practice: The role of self concept in educational psychology. Leicester, Nicholson, J. S., Deboeck, P. R., & Howard, W. (2017). Attrition in
UK: British Psychological Society. developmental psychology: A review of modern missing data reporting
and practices. International Journal of Behavioral Development, 41, achievement settings. In S. Karabenick & T. C. Urdan (Eds.), Advances
143–153. http://dx.doi.org/10.1177/0165025415618275 in motivation and achievement (Vol. 18, pp. 1–35). Bingley, UK: Em-
O’Connor, R. E., Beach, K., Sanchez, V., Bocian, K., & Flynn, L. erald Group Publishing Limited. http://dx.doi.org/10.1108/S0749-
(2015). Building BRIDGES: A design experiment to improve reading 742320140000018000
and United States history knowledge of poor readers in 8th grade. Putwain, D. W., & Aveyard, B. (2018). Is perceived control a critical factor
Exceptional Children, 81, 399 – 425. http://dx.doi.org/10.1177/ in understanding the negative relationship between cognitive test anxiety
0014402914563706 and examination performance? School Psychology Quarterly, 33, 65–74.
Peixoto, F., Sanches, C., Mata, L., & Monteiro, V. (2017). “How do you http://dx.doi.org/10.1037/spq0000183
feel about math?”: Relationships between competence and value ap- Putwain, D. W., Becker, S., Symes, W., & Pekrun, R. (2018). Reciprocal
praisals, achievement emotions and academic achievement. European relations between students’ academic enjoyment, boredom, and achieve-
Journal of Psychology of Education, 32, 385– 405. http://dx.doi.org/10 ment over time. Learning and Instruction, 54, 73– 81. http://dx.doi.org/
.1007/s10212-016-0299-4 10.1016/j.learninstruc.2017.08.004
Pekrun, R. (2006). The control-value theory of achievement emotions: Putwain, D. W., Connors, L., Symes, W., & Douglas-Osborn, E. (2012). Is
Assumptions, corollaries, and implications for educational research and academic buoyancy anything more than adaptive coping? Anxiety,
practice. Educational Psychology Review, 18, 315–341. http://dx.doi Stress, & Coping, 25, 349 –358. http://dx.doi.org/10.1080/10615806
.org/10.1007/s10648-006-9029-9 .2011.582459
Pekrun, R. (2016). Academic emotions. In K. R. Wentzel & D. B. Miele Putwain, D. W., Daly, A. L., Chamberlain, S., & Sadreddini, S. (2015).
(Eds.), Handbook of motivation in school (2nd, pp. 120 –145). New Academically buoyant students are less anxious about and perform
York, NY: Routlegde. better in high-stakes examinations. British Journal of Educational Psy-
Pekrun, R. (2017). Achievement emotions. In A. J. Ellior, C. S. Dweck, & chology, 85, 247–263. http://dx.doi.org/10.1111/bjep.12068
D. S. yeager (Eds.) Handbook of competence and motivation (2nd ed., Putwain, D. W., Daly, T., Chamberlain, S., & Saddredini, S. (2016). “Sink
pp. 251–271). New York, NY: Guilford Press. or swim”: Buoyancy and coping in the cognitive test anxiety–academic
Pekrun, R. (2018). Control-value theory: A social-cognitive approach to performance relationship. Educational Psychology, 36, 1807–1825.
achievement emotions. In G. A. D. Liem & D. M. McInerney (Eds.), Big http://dx.doi.org/10.1080/01443410.2015.1066493
theories revisited 2: A volume of research on sociocultural influences on Putwain, D. W., Schmitz, E. A., Wood, P., & Pekrun, R. (2020). The role
motivation and learning (pp. 162–190). Charlotte, NC: Information Age
of achievement emotions in primary school mathematics: Control-value
Publishing.
antecedents and achievement outcomes. British Journal of Educational
Pekrun, R. (in press). Self-appraisals and emotions: A control-value ap-
Psychology. Advance online publication. http://dx.doi.org/10.1111/bjep
proach. In T. Dicke, F. Guay, H. W. Marsh, R. G. Craven, & D. M.
.12367
McInerney (Eds.), Self: A multidisciplinary concept. Charlotte, NC:
Raccanello, D., Brondino, M., Moe, A., Stupnisky, R., & Lichtenfeld, S.
Information Age Publishing.
(2019). Enjoyment, boredom, and anxiety in elementary schools in two
Pekrun, R., Goetz, T., Frenzel, A. C., Barchfeld, P., & Perry, R. P. (2011).
domains: Relations with achievement. Journal of Experimental Educa-
Measuring emotions in students’ learning and performance: The
tion, 87, 449 – 469. http://dx.doi.org/10.1080/00220973.2018.1448747
Achievement Emotions Questionnaire (AEQ). Contemporary Educa-
Ramirez, G., Shaw, S. T., & Maloney, E. A. (2018). Math anxiety: Past
tional Psychology, 36, 36 – 48. http://dx.doi.org/10.1016/j.cedpsych
research, promising interventions, and a new interpretation framework.
.2010.10.002
Educational Psychologist, 53, 145–164. http://dx.doi.org/10.1080/
Pekrun, R., Goetz, T., Titz, W., & Perry, R. P. (2002). Academic emotions
00461520.2018.1447384
in students’ self-regulated learning and achievement: A program of
Rosel, J., & Plewis, I. (2008). Longitudinal data analysis with structural
qualitative and quantitative research. Educational Psychologist, 37, 91–
105. http://dx.doi.org/10.1207/S15326985EP3702_4 equations. Methodology, 4, 37–50. http://dx.doi.org/10.1027/1614-2241
Pekrun, R., Hall, N. C., Goetz, T., & Perry, R. P. (2014). Boredom and .4.1.37
academic achievement: Testing a model of reciprocal causation. Journal Salles, A., Ais, J., Semelman, M., Sigman, M., & Calero, C. I. (2016). The
of Educational Psychology, 106, 696 –710. http://dx.doi.org/10.1037/ metacognitive abilities of children and adults. Cognitive Development,
a0036006 40, 101–110. http://dx.doi.org/10.1016/j.cogdev.2016.08.009
Pekrun, R., Lichtenfeld, S., Marsh, H. W., Murayama, K., & Goetz, T. Schaeffer, M. W., Rozek, C. S., Berkowitz, T., Levine, S. C., & Beilock,
(2017). Achievement emotions and academic performance: Longitudinal S. L. (2018). Disassociating the relation between parents’ math anxiety
models of reciprocal effects. Child Development, 88, 1653–1670. http:// and children’s math achievement: Long-term effects of a math app
dx.doi.org/10.1111/cdev.12704 intervention. Journal of Experimental Psychology: General, 147, 1782–
Pekrun, R., & Loderer, K. (in press). Control-value theory and students 1790. http://dx.doi.org/10.1037/xge0000490
with special needs: Achievement emotion disorders and their links to Shao, K., Pekrun, R., Marsh, H. W., & Loderer, K. (2020). Control-value
behavioral disorders and academic difficulties. In A. J. Martin, R. A. appraisals, achievement emotions, and foreign language performance: A
Sperling, & K. J. Newton (Eds.), Handbook of educational psychology latent interaction analysis. Learning and Instruction, 69, 101356. http://
and students with special needs. London, UK: Taylor & Francis. dx.doi.org/10.1016/j.learninstruc.2020.101356
Pekrun, R., Murayama, K., Marsh, H. W., Goetz, T., & Frenzel, A. C. Sinatra, G. M., Broughton, S. H., & Lombardi, D. (2016). Emotions in
(2019). Happy fish in little ponds: Testing a reference group model of science education. In R. Pekrun & L. Linnenbrink-Garcia (Eds.), Inter-
achievement and emotion. Journal of Personality and Social Psychol- national handbook of emotions in education (pp. 415– 436). London,
ogy, 117, 166 –185. http://dx.doi.org/10.1037/pspp0000230 UK: Routledge.
Pekrun, R., & Perry, R. P. (2014). Control-value theory of achievement Smith, T. F., & Capuzzi, G. (2019). Using a mindset intervention to reduce anxiety
emotions. In R. Pekrun & L. Linnenbrink-Garcia (Eds.), International in the statistics classroom. Psychology Teaching and Learning. Advance online
handbook of emotions in education (pp. 120 –141). London, UK: Rout- publication. http://dx.doi.org/10.1177/1475725719836641
ledge. http://dx.doi.org/10.4324/9780203148211 Standards and Testing Agency. (2016a). 2016 National Curriculum Tests,
Perry, R. P., Chipperfield, J. G., Hladkyj, S., Pekrun, R., & Hamm, Key Stage 2 Mathematics Paper 2: Reasoning. Richmond, UK: Her
J. M. (2014). Attribution-based treatment interventions in some Majesty’s Stationery Office.
Standards and Testing Agency. (2016b). 2016 National Curriculum Tests, treatment studies from 2000 –2010. Psychology in the Schools, 50,
Key Stage 2 Mathematics Paper 3: Reasoning. Richmond, UK: Her 57–71. http://dx.doi.org/10.1002/pits.21660
Majesty’s Stationery Office. von der Embse, N., Jester, D., Roy, D., & Post, J. (2018). Test anxiety
Standards and Testing Agency. (2017a). 2017 National Curriculum Tests, effects, predictors, and correlates: A 30-year meta-analytic review. Jour-
Key Stage 2 Mathematics Paper 2: Reasoning. Richmond, UK: Her nal of Affective Disorders, 227, 483– 493. http://dx.doi.org/10.1016/j.jad
Majesty’s Stationery Office. .2017.11.048
Standards and Testing Agency. (2017b). 2017 National Curriculum Tests, von der Embse, N. P., & Witmer, S. (2014). High-stakes accountability:
Key Stage 2 Mathematics Paper 3: Reasoning. Richmond, UK: Her Student anxiety and large-scale testing. Journal of Applied School Psy-
Majesty’s Stationery Office. chology, 30, 132–156. http://dx.doi.org/10.1080/15377903.2014.888529
Standards and Testing Agency. (2018a). 2018 National Curriculum Tests, Warren, M. K., Ollendick, T. H., & King, N. J. (1996). Test anxiety in girls
Key Stage 2 Mathematics Paper 2: Reasoning. Richmond, UK: Her and boys: A clinical-developmental analysis. Behaviour Change, 13,
Majesty’s Stationery Office. 157–170. http://dx.doi.org/10.1017/S0813483900004939
Standards and Testing Agency. (2018b). 2018 National Curriculum Tests, Watkins, M. W., Lei, P. W., & Canivez, G. L. (2007). Psychometric
intelligence and achievement: A cross-lagged panel analysis. Intelli-
Key Stage 2 Mathematics Paper 3: Reasoning. Richmond, UK: Her
gence, 35, 59 – 68. http://dx.doi.org/10.1016/j.intell.2006.04.005

Majesty’s Stationery Office.
Widaman, K. F., Ferrer, E., & Conger, R. D. (2010). Factorial invariance
Swann, W. B., Jr., Chang-Schneider, C., & Larsen McClarty, K. (2007).

within longitudinal structural equation models: Measuring the same
Do people’s self-views matter? Self-concept and self-esteem in everyday
construct across time. Child Development Perspectives, 4, 10 –18. http://
life. American Psychologist, 62, 84 –94. http://dx.doi.org/10.1037/0003-
dx.doi.org/10.1111/j.1750-8606.2009.00110.x
066X.62.2.84
Wigfield, A. (1997). Reading motivation: A domain specific approach to
Symes, W., Putwain, D. W., & Remedios, R. (2015). Academic buoyancy motivation. Educational Psychologist, 32, 59 – 68. http://dx.doi.org/10
and the appraisal of fear appeals, used prior to a high-stakes examina- .1207/s15326985ep3202_1
tion, as threatening or challenging. School Psychology International, 36, Wu, J.-W., & Kwok, O.-W. (2012). Using SEM to analyze complex survey
605– 619. http://dx.doi.org/10.1177/0143034315610622 data: A Comparison between design-based single-level and model-based
Tze, V. M., Daniels, L. M., & Klassen, R. M. (2016). Evaluating the multilevel approaches. Structural Equation Modeling, 19, 16 –35. http://
relationship between boredom and academic outcomes: A meta-analysis. dx.doi.org/10.1080/10705511.2012.634703
Educational Psychology Review, 28, 119 –144. http://dx.doi.org/10 Yun, S., Hiver, P., & Al-Hoorie, A. H. (2018). Academic buoyancy:
.1007/s10648-015-9301-y Exploring learners’ everyday resilience in the language classroom. Stud-
Villavicencio, F. T., & Bernardo, A. B. I. (2016). Beyond math anxiety: ies in Second Language Acquisition, 40, 805– 830. http://dx.doi.org/10
Positive emotions predict mathematics achievement, self-regulation and .1017/S0272263118000037
self-efficacy. The Asia-Pacific Education Researcher, 25, 415– 422.
http://dx.doi.org/10.1007/s40299-015-0251-4 Received March 10, 2020
von der Embse, N., Barterian, J., & Segool, N. (2013). Test anxiety Revision received August 1, 2020
interventions for children and adolescents: A systematic review of Accepted August 6, 2020 䡲
E-Mail Notification of Your Latest Issue Online!

Would you like to know when the next issue of your favorite APA journal will be available
online? This service is now available to you. Sign up at https://my.apa.org/portal/alerts/ and you will
be notified by e-mail when issues of interest to you become available!

Putwain2022 (PEDIR POR SEXO)

Uploaded by

Copyright:

Available Formats

Putwain2022 (PEDIR POR SEXO)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Putwain2022 (PEDIR POR SEXO)

Uploaded by

Copyright:

Available Formats

Journal of Educational Psychology

© 2020 American Psychological Association 2022, Vol. 114, No. 1, 108–126

Achievement Emotions and Academic Achievement: Reciprocal Relations

David W. Putwain and Peter Wood Reinhard Pekrun

Educational Impact and Implications Statement

Keywords: academic achievement, achievement emotions, anxiety, buoyancy, control-value theory

interaction between academic buoyancy and boredom or anxiety

would therefore be expected. At low boredom or anxiety, there

relations between achievement emotions and achievement, with ure 1):

T1 Boredom T2 Maths Test T3 Boredom T4 Maths Test

collection. The script required teachers to check the voluntary

participation of each student, verbally, and provide an alternative

Item M SD ␣/␻ ␳I Skewness Kurtosis Factor loadings

T1 Enjoyment 16.24 4.48 .92/.92 .07 ⫺1.09 0.21 .84–.90

which, for mathematics, at Key Stage 2 covers arithmetic, mea-

Variable ␹2(df) RMSEA SRMR CFI TLI ⌬RMSEA ⌬CFI ⌬TLI

Model ␹2(df) RMSEA SRMR CFI TLI AIC ⌬AIC TRd(df)

T1 Enjoyment .59 T3 Enjoyment

T1 Boredom T2 Maths Test -.14 T3 Boredom T4 Maths Test

T1 Anxiety .50 T3 Anxiety

Enjoyment (ENJ) .04 (.06)

T4 Mathematics Test Performance

gence, 35, 59 – 68. http://dx.doi.org/10.1016/j.intell.2006.04.005

Swann, W. B., Jr., Chang-Schneider, C., & Larsen McClarty, K. (2007).

E-Mail Notification of Your Latest Issue Online!

You might also like