-
1
School attachment and relatedness with parents, friends and
teachers
as predictors of students’ intrinsic and identified
regulation
Frédéric Guay, Anne-Sophie Denault, Stéphanie Renaud Université
Laval, Faculty of Education, Canada
Abstract This longitudinal study examined the role of school
attachment and three sources of relatedness (friends, parents,
teachers) in predicting students’ intrinsic and identified
regulation. A total of 946 high school students from disadvantaged
neighborhoods completed self-report measures. Results of a CFA
provided support for the distinctiveness of the sources of
relatedness and school attachment. Results of SEM revealed that
school attachment predicted intrinsic regulation, whereas
relatedness to teachers predicted identified regulation.
Relatedness to parents and friends did not explain a significant
percentage of the variance in outcomes. All results were obtained
while controlling for initial levels of the outcomes. In sum,
students with low levels of intrinsic and identified regulation for
learning activities may benefit from practices designed to increase
school attachment and relatedness with their teachers.
1. Introduction Positive psychology stresses the importance of
understanding what makes people happy with
their lives, from birth to death (Peterson, 2006), in numerous
life contexts, including school (Huebner, Gilman, & Furlong,
2009). Intrinsic and identified regulation for learning activities
are two positive educational characteristics that foster optimal
functioning at school (Ryan & Deci, 2009) and that are affected
by the quality of emotional bonds with significant others. More
specifically, when students feel attached to their school and
experience relatedness (i.e., close and secure emotional
relationships; Deci & Ryan, 2000) with teachers, parents and
friends, they derive greater pleasure from school-based learning
activities (intrinsic regulation) and are also more likely to
internalize the value of these activities (identified
regulation).
However, few studies to date have simultaneously investigated
multiple sources of relatedness (i.e., parents, teachers and
friends) in predicting intrinsic and identified regulation for
school-related activities among high school students. Examining
various sources of relatedness is important because, if one source
is found to be more relevant than another for fostering these
outcomes, intervention strategies could be designed to focus on
that specific source. Alternatively, if all sources of relatedness
are found to be important, this could lead to systemic intervention
strategies. Moreover, while the concept of school attachment has
been the focus of several studies over the past decades, much of
the literature on this subject comes from sociology and community
psychology and includes very few comparisons with the family,
classroom, or friendship contexts (Hill & Werner, 2006). No
study to date has established whether these sources of relatedness
and school attachment are interdependent, something that would
further our understanding of intrinsic and identified regulation
for learning activities. In addition, one of the dominant
perspectives in the field of motivation posits that both
relatedness with others and school attachment improve intrinsic and
identified regulation for learning activities. Yet, the opposite
perspective, positing that students’ motivational beliefs may
affect their perceptions of the quality of their relationships
-
2
with others and school attachment, has rarely been tested (see
Guay, Marsh, Senécal, & Dowson, 2008 for an exception).
Moreover, as proposed by some gendered models of socialization
(Gilligan, 1982; Maccoby, 1998), girls are more likely than boys to
value and report higher levels of relatedness with significant
others (e.g., parents, teachers and friends) and school attachment.
However, less is known about gender differences in the predictive
associations between these sources of relatedness and school
attachment and intrinsic and identified regulation. For example,
while we might expect adolescent girls to report greater
relatedness with their teachers than boys, it is not clear whether
relatedness with teachers will predict higher motivational outcomes
among girls than among boys. Looking at such gender differences is
important because it could lead to the development of optimal
gender-specific intervention strategies. Finally, students in
disadvantaged neighborhoods may have trouble mobilizing their
motivation at school. They may have restricted opportunities
(Schoon, 2008). For example, their teachers might be experiencing
particularly high levels of stress because they face challenges
such as a higher suspension rate (O’Brennan, Pas, & Bradshaw,
2017) and academic difficulties (Fantuzzo et al., 1999) leading
them to use less optimal pedagogical practices. In addition,
parents work schedules in these contexts may be more inflexible
compared to more affluent populations, precluding them from being
actively involved in the homework setting (Cooper, Lindsay, &
Nye, 2000). Moreover, students themsleves and their friends may not
view school as a viable pathway to success (Oyserman, 2013).
Investigating interpersonal/school factors among students living in
disadvantaged neighborhoods could lead to a better understanding of
what makes them resilient (Battistich et al., 1995). Taken
together, there is a need to better investigate the relationships
between school attachment, sources of relatedness and types of
motivation for both girls and boys, and also for students who are
at risk of developing low motivation due to limited neighborhood
resources.
Accordingly, four goals were pursued in this longitudinal study.
First, we investigated the potential effects of three sources of
relatedness (parents, teachers and friends) and school attachment
on intrinsic and identified regulation for school-based learning
activities among high school students living in disadvantaged
neighborhoods. Second, we examined whether these sources of
relatedness and school attachment were interdependent. For example,
students who have positive relationships with their teachers might
be more highly valued by their friends at school. Third, we
investigated the reciprocal associations between the three sources
of relatedness and school attachment, and intrinsic and identified
regulation for school-based learning activities. Fourth and lastly,
we tested for gender differences in the associations between the
three sources of relatedness and school attachment and intrinsic
and identified regulation for school-based learning activities.
1.1. Intrinsic and identified regulation Intrinsic motivation is
defined as engaging in an activity for its own sake—that is,
because the
activity is interesting and enjoyable, and when there are no
extrinsic motivators (Ryan & Deci, 2009). Intrinsically
motivated behaviors are archetypes of autonomy. Intrinsic
motivation is considered to be inherent in human nature; it is the
manifestation of people’s innate proactivity. This is evident, for
example, when adolescents are engaged and enthusiastic in a
responsive school environment. Identified regulation, for its part,
occurs when extrinsically motivated behaviors are performed with a
sense of choice and volition, such as when students truly consider
behaviors to be important for their own personal development.
Individuals identify with the importance and value of
self-regulating their behaviors, which then become part of their
own values (Deci & Ryan,
-
3
2000). Although these behaviors are considered to be
instrumental, they nevertheless become integrated into the
individuals’ sense of self, to the point where they are performed
autonomously. These two types of regulation are differentially
associated with a host of positive outcomes in adolescence, such as
achievement and persistence (see Ryan & Deci, 2009). For
example, Burton et al. (2006) showed that identified motivation
predicted academic achievement whereas intrinsic motivation
predicted psychological well-being among students.
Self-determination theory (SDT) also posits that relatedness is
an essential prerequisite for the initiation of behaviors driven by
intrinsic or identified motivation (Deci & Ryan, 1991).
Relatedness is especially important when students are not
intrinsically motivated to attend high school but nevertheless need
to endorse school values (e.g., studying for exams). In this case,
the internalization of external requirements is more likely to
occur when students feel connected to significant others (Deci
& Ryan, 1991), especially if students come from disadvantaged
neighborhoods where they might have experienced various
developmental challenges (Burney & Beilke, 2008). In contrast,
when students feel detached from others, these external
requirements have less chance of being integrated into the self.
SDT thus hypothesizes that relatedness with significant others is
positively related to intrinsic and identified regulation.
1.2. Various sources of relatedness, school attachment and
intrinsic and identified
regulation
Relatedness is considered a basic human need in many theories
(see Baumeister & Leary, 1995). Relatedness is defined within
SDT as the basic psychological need to feel connected to others, to
love and feel loved, to care and feel cared for (Deci & Ryan,
2000). Before reviewing studies on the associations between various
sources of relatedness and intrinsic and identified regulation, it
is important to describe more generally how these sources might
impact students’ development during adolescence. Two different
perspectives appear to emerge from the literature. The first
suggests that, as adolescents grow older, they rely less on their
parents or family (Scholte & van Aken, 2006). Theories such as
those based on the neo-psychoanalytic, evolutionary or socio-
cognitive perspective endorse this point of view, positing that
increasing autonomy and individuation in adolescence cause
closeness with parents to temporarily decrease, conflicts to
increase, and power to progressively equalize (De Goede, Branje,
& Meeus, 2008). According to this perspective, detachment sets
the stage for greater self-direction and the possibility to
establish new significant relationships, such as those with peers.
The second perspective, on the other hand, posits that the process
of separation–individuation does not occur at the expense of
relatedness with parents (Youniss & Smollar, 1985). According
to this perspective, attachment to parents offers the possibility
to develop autonomy and establish new significant relationships
with others, such as friends and teachers (Ryan, 1993). In other
words, relatedness with parents and other sources of relatedness
are mutually supportive.
This theoretical background could lead to different hypotheses
regarding the role of each
source of relatedness in predicting intrinsic and extrinsic
regulation. The first perspective would posit that relatedness with
parents is less important during adolescence in terms of supporting
these motivational resources whereas the second would suggest that
all sources of relatedness are important. These two perspectives
can also be tested in light of the additive and threshold models
(see Laursen & Mooney, 2008). According to the additive model,
intrinsic and identified regulation for learning activities could
be cumulatively affected by all significant relationships that
fulfill the need for relatedness. The additive model is in line
with the sensitization-desensitization
-
4
perspective (Moller et al., 2010) according to which people who
experience relatedness satisfaction in one relationship continue to
value its satisfaction in other relationships.
In contrast, the threshold model assumes that individuals do not
need to perceive relatedness
in various relationships and that relational support is
redundant, such that additional sources of relatedness do not
improve psychological adjustment. This last perspective is in line
with the satiated perspective regarding psychological needs,
whereby individuals who have satisfied their need for relatedness
in one relationship will not experience further benefits from other
positive relationships (Baumeister & Leary, 1995). For example,
students who have satisfied their need for relatedness with their
peers might assign less value to time spent with their parents or
teachers because their need for relatedness has already been
met.
A limited number of studies have contrasted the effects of
various sources of relatedness on
student outcomes. For example, Ryan, Stiller, and Lynch (1994)
reported that adolescents’ relatedness with both parents and
teachers were significantly associated with intrinsic and
identified regulation, while their relatedness with friends was
not. However, Furrer and Skinner (2003) showed that a sense of
relatedness to parents, teachers and peers each predicted students’
emotional and behavioral engagement. Indeed, they found that
self-reported emotional and behavioral engagement were highest when
children perceived high relatedness with all these significant
sources of relatedness. In a study among college students, Guay et
al. (2008) tested a cross-lagged model in which they contrasted two
sources of relatedness: parents and friends. Their results
indicated that only relatedness with parents predicted subsequent
intrinsic and identified regulation. However, their study did not
include relatedness with teachers, contrary to the other two
studies. Thus, among studies investigating such additive or
threshold effects, no clear picture has emerged regarding the
importance of one source of relatedness over the other. The
question as to whether one source of relatedness is more important
than the others during adolescence thus remains open, although some
recent studies in social psychology have revealed the incremental
value of relatedness experiences (Moller et al., 2010).
In addition, previous studies have not included school
attachment, which is an emotional
feeling of affection for and enjoyment of school (Goodenow,
1993). Although conceptually related to the relatedness construct
(i.e., trust and satisfaction), school attachment is different
insofar as it is not evaluated in reference to a specific
interpersonal relationship but rather toward school in general and
thus represents a general feeling about the school experience.
Moreover, it should not be confused with similar terms such as
school connectedness (Niehaus et al., 2012), which also refers to
specific relationships at school, or school bonding, based on a
conglomerate of interpersonal experiences at school. In the current
study, we used a measure that clearly assessed the students’
emotional feeling of affection for and enjoyment of school. School
attachment has been linked to a host of positive outcomes,
including positive social, emotional and academic adjustment.
Numerous interventions have demonstrated that improving the school
climate in high schools results in greater school attachment and
other desirable outcomes, such as lower rates of delinquency,
substance abuse and school problems (Hill & Werner, 2006).
Regarding indicators of motivation, school attachment is
significantly and positively associated with expectancy of success,
valuing schoolwork, general school motivation, and self-reported
effort (Goodenow & Grady, 1993), as well as self-efficacy and
self-concept (Chiu, Chow, McBride, & Mol, 2016). As an
important predictor of many positive outcomes, school attachment
should be included among the possible predictors of intrinsic and
identified regulation. In fact, school attachment may be more
important than sources of relatedness in predicting these forms of
regulation. However, to
-
5
test this assertion rigorously, it is important to consider
possible positive interdependencies among sources of relatedness
and school attachment.
1.3. Positive interdependencies among sources of relatedness and
school attachment The fact that various sources of relatedness and
school attachment might influence one another
represents a challenge when analyzing their effects in an
additive or threshold model. More specifically, if there proves to
be no predictive association between relatedness to teachers and
intrinsic and identified regulation, it might nevertheless be the
case that relatedness to teachers has an indirect effect on these
forms of regulation through school attachment (Anderman, 2003; Chiu
et al., 2016). For example, students who perceive a great deal of
support from their teachers might experience a higher sense of
school attachment. Similarly, parents who have a positive
relationship with their children may help them develop a generally
positive attitude toward school, thus fostering greater school
attachment. In addition, students who have a positive relationship
with their teachers are more likely to be accepted by their peers
(Davidson, Gest, & Welsh, 2010), which suggests that students
pay attention to their relationships with the teacher and use this
information in establishing positive relationships with friends. In
sum, there might be many positive interdependencies among sources
of relatedness and school attachment that need to be accounted for
when examining their unique contribution to intrinsic and
identified regulation.
1.4. Reciprocal relationships between sources of relatedness,
school attachment and motivation
Most researchers endorse the view that sources of relatedness
and school attachment “predict” intrinsic or identified regulation
(Guay et al., 2008). However, as argued earlier, it is also
possible that relatedness with significant others and school
attachment are reinforced by the fact that students are motivated
for intrinsic or identified reasons. Indeed, adolescents and young
adults are not passive recipients of the social context. They are
active agents, evoking responses from both parents and friends,
which may, in turn, modify their sense of relatedness with
significant others (Guay et al., 2008). For example, a student may
enjoy learning activities and consider them to be important, which
may subsequently affect his/her behaviors (e.g., seeking support
from friends, studying more at home, being more engaged at school).
In turn, his/her own behaviors might elicit behaviors from friends
(e.g., giving support), and parents and teachers (e.g., giving
positive feed- back), thereby leading to better representations of
relatedness with these individuals. In addition, intrinsic and
identified regulation for school work is likely to make one feel
much more attached to school. Guay et al. (2008) tested these
reciprocal relations. Their results provided good support for the
effect of relatedness with parents on autonomous academic
motivation but no convincing support for the effect of motivation
on relatedness with parents. In addition, no significant effect was
found in either direction between relatedness with friends and
autonomous academic motivation. However, their study involved
college students and did not consider school attachment or
relatedness with teachers. Thus, it is not known whether there are
reciprocal associations between these latter variables and
intrinsic and identified regulation among high school students.
However, based on SDT, one would expect school attachment and
relatedness with teachers to be predictors of intrinsic and
identified regulation rather than the reverse.
1.5. Moderating role of gender
-
6
Gender differences are usually considered to be important when
it comes to interpersonal relationships. As suggested by some
developmental theories on gender differences (e.g., Gilligan, 1982;
Maccoby, 1998), the differences in how girls and boys are
socialized concerning social relationships lead girls to value
relatedness and invest more time in their social relationships than
boys. In addition, according to the gender intensification theory,
this may be especially true during adolescence when the physical
changes of puberty increase the pressure for sex-typed behaviors
among individuals (Hill & Lynch, 1983). Moreover, girls usually
value close relationships for their emotional and expressive
qualities while boys value them for their instrumental features
(Burleson, 2003). In the same vein, gender differences in mean
levels of perceived support have been consistently documented in
prior studies on concepts that are similar to relatedness. For
example, in a study examining gender differences in the association
between perceived social support and attitude toward school among
adolescents, even though boys and girls reported the same levels of
support from parents, compared to boys, girls reported
significantly higher levels of support from teachers, classmates,
close friends and school personnel (Rueger, Malecki, & Demaray,
2010).
Yet, the question remains as to whether boys and girls benefit
equally from relatedness or school attachment. In other words, does
perceiving greater relatedness with significant others have a
greater tendency to increase levels of intrinsic and identified
regulation among either girls or boys? SDT proposes that
relatedness is a fundamental need that should provide the essential
nutriments for all individuals, including both boys and girls, to
develop (Deci & Ryan, 2000). Without this need it would be hard
to explain why girls and boys would so readily internalize ways of
interacting effectively and harmoniously with others in their peer
groups. Satisfaction of the fundamental need for relatedness also
helps boys and girls develop intrinsic and identified regulation
(Vallerand, Fortier, & Guay, 1997). For example, intrinsic and
identified regulation are likely to emerge in a school environment
where boys and girls feel unconditional regard and support,
allowing them to focus on their skills. Results from two
longitudinal studies supported this perspective. Specifically,
these studies revealed no gender differences in the associations
between various sources of social support (e.g., parents, teachers,
peers) and school outcomes (i.e., attitude toward school and
engagement) (Rueger et al., 2010; Wang & Eccles, 2012). In line
with these results, despite the existence of gender differences
pertaining to some sources of relatedness or support, we believe
that the functions served by these sources are more similar than
different across gender. Thus, males and females are likely to have
similar conceptions of close relationships and similar expectations
regarding the sense of relatedness that these relationships provide
(Burleson, 2003).
1.6. Overview of the present study
Using an autoregressive cross-lagged design with two measurement
times, we pursued the
following four goals: First, we investigated the unique
predictive associations between three sources of relatedness
(parents, teachers and friends) and school attachment, on the one
hand, and intrinsic and identified regulation for school-based
learning activities, on the other. Based on SDT postulates
regarding the nature and functions of the need for relatedness, we
hypothesized that all three sources of relatedness and school
attachment would be important for predicting intrinsic and
identified regulation (the additive model). Second, we examined
whether the sources of relatedness and school attachment were
interdependent. In line with previous findings, we expected some
interdependent effects to emerge, for example, relatedness with
teachers leading to greater school attachment. Third, we
investigated reciprocal associations between the sources of
relatedness (parents, friends and teachers) and school attachment
and intrinsic and identified regulation. Based
-
7
on SDT and previous studies, we expected sources of relatedness
and school attachment to predict intrinsic and identified
regulation rather than the reverse. Fourth, we tested the
moderating effect of gender. Based on theory and previous findings,
no significant gender differences were expected at the process
level, in contrast to the mean level. Mean-level differences refer
to variances in a dependent variable between boys and girls,
whereas process-level differences indicate a dissimilarity in the
degree of association between an independent and dependent variable
among boys and girls. Though we expected no difference at the
process-level, we expected the following differences at the mean
level based on Rueger et al. (2010) and Vallerand et al. (1997):
comparatively to boys, girls will report higher levels of
relatedness with friends and teachers, school attachment as well as
intrinsic and identified regulations. These questions were examined
among a sample of high school students from disadvantaged
neighborhoods. These students are usually considered to be at
greater risk of school dropout (Donnelly, 2015). Students from
disadvantaged backgrounds are less likely to perceive their
relationships with others as positive (Crosnoe et al., 2004). Thus,
sources of relatedness and school attachment might be especially
important for these students’ academic outcomes such as intrinsic
and identified regulation (see Rudasill, Niehaus, Crockett, &
Rakes, 2014).
2. Method
2.1. Participants, study design and procedures
The participants took part in a longitudinal research project
examining extracurricular
activities and school dropout among students from disadvantaged
neighborhoods. In this research project, a stratified sample of
3000 students in Grades 7 to 10 from disadvantaged neighborhoods in
the province of Quebec (Canada) was formed based on sex, grade
level and administrative region. To be included in the sample,
students had to attend schools with a score of 8, 9 or 10 on the
two indices of deprivation provided by the MELS (2007–2008).
According to these indices, schools are rated on a scale from 1 to
10, where 1 is considered the most advantaged and 10, the most
disadvantaged. We asked the ministry of education to provide a
sample of 3000 students in order to have a large enough sample on
which to perform structural equation analyses. Based on previous
research (Richards et al., 2010), we estimated that approximately
33% of students would agree to take part in the study.
A questionnaire was thus mailed to the 3000 students in the
stratified sample and 952 (32%) of them completed and returned the
questionnaires (T1; 56% girls). One year later (T2), 639 of these
students completed and returned another questionnaire (67%
retention rate; 58% girls). Differences in T1 socio-demographic
information between completers (i.e., those who participated at
both time points) and partial completers (i.e., those who
participated only at T1) revealed that the completers were more
likely to be girls, younger, and from intact and more educated
families than the partial completers. We conducted a series of
invariance analyses to determine whether the statistical parameters
(factor loadings, intercepts, uniquenesses, variances and
covariances, latent means) differed between completers and partial
completers. The results are presented in the data analysis
section.
2.2. Measures
2.2.1. Intrinsic and identified regulation
-
8
We assessed intrinsic and identified academic regulation using
the French version of the Academic Motivation Scale (AMS;
Vallerand, Blais, Briere, & Pelletier, 1989). The AMS includes
seven subscales, each containing four items representing a possible
reason (or regulation) for attending school. Three subscales assess
types of intrinsic motivation: intrinsic motivation to know, to
accomplish things, and to experience stimulation. Three other
scales assess types of extrinsic motivation: identified,
introjected and external regulation. The seventh subscale assesses
amotivation. Items are scored on a five-point Likert scale ranging
from 1 (strongly disagree) to 5 (strongly agree). Numerous studies
have supported the factorial, convergent and divergent validity and
internal consistency of the AMS (Vallerand et al., 1989, 1992,
1993).
As mentioned earlier, we used only the following subscales:
Intrinsic Motivation to Know (i.e. intrinsic regulation) (e.g.,
Because I experience pleasure and satisfaction while learning new
things) and Identified Regulation (e.g., Because eventually it will
enable me to enter the job market in a field that I like).
Correlations between these two subscales (see Table 4) were
moderate and consistent with past research (e.g., Ratelle, Guay,
Vallerand, Larose, & Senécal, 2007). Table 1 presents means,
standard deviations, skewness and kurtosis, and minimum and maximum
values for each item. Based on criteria formulated by In’nami and
Koizumi (2003), all skewness (−2 to +2 range) and kurtosis
(acceptable range −7 to +7) values were acceptable, except for the
kurtosis values for the second identified regulation item (7.32 at
T1, 10.16 at T2). This item was nevertheless retained in the
analyses because the departure from the criteria values was not
very high. Scale score reliability estimates were computed using
the standardized parameter estimates from the confirmatory factor
analysis (CFA), using McDonald’s (1970) formula:
ω=(Σ|λi|)2/([Σ|λi|]2+Σδii), where λi represents the standardized
factor loadings and δii represents the standardized item
uniquenesses. Compared to traditional scale score reliability
estimates (e.g., alpha), ω has the advantage of taking into account
the strength of association between items and constructs (λi) as
well as item-specific measurement errors (δii). All four scale
score reliability estimates were 0.80 for identified and intrinsic
regulation at both time points.
2.2.2. Sources of relatedness Relatedness with parents, teachers
and friends, respectively, was measured using the French
version of the Interpersonal Relationships Quality Scale
developed by Senécal, Vallerand, and Vallières (1992). This scale
includes four items (e.g., “My relations with this person are
satisfying”; “My relations with this person are trustworthy”) which
students are asked to answer in reference to each source of
relatedness, namely teachers, friends and parents. Items are scored
on a five-point Likert scale ranging from 0 (strongly disagree) to
4 (strongly agree). Senécal et al.’s (1992) study supported the
validity and internal consistency of this scale. In our study, the
McDonald’s ω ranged between 0.79 and 0.80 for all sources at both
time points. All descriptive values were acceptable (i.e., no floor
or ceiling effects, sufficient variability, adequate skewness and
kurtosis; see Table 1).
-
9
-
10
2.2.3. School attachment
We used a five-item measure adapted by Hill and Werner (2006)
from three instruments developed respectively by Eccles et al.
(1993), Hawkins, Guo, Hill, Battin-Pearson, and Abbott (2001) and
Barber and Olsen (1997). These five items are: “I am proud to be at
this school”; “I am happy to be at this school”; “I feel safe in my
school”; “Most mornings I look forward to going to school”; and “I
like my school.” Items are scored on a five-point Likert scale
ranging from 1 (strongly disagree) to 5 (strongly agree). In their
study, Hill and Werner reported factor loadings ranging from 0.69
to 0.89 for these items. Also, in support of the construct validity
of this measure, school attachment was negatively correlated with
aggressive behaviors at school. Specifically, school attachment is
expected to indicate that affiliative goals are met and increase
behavioral engagement in school, which should result in less
aggression in schools (Hill & Werner, 2006). In our study, the
McDonald’s ω was 0.83 at T1 and 0.82 at T2. All descriptive values
were acceptable (see Table 1).
2.3. Data analysis
All models were estimated using Mplus (Version 7.4; Muthén &
Muthén, 2012) and tested
using standardized coefficients obtained through the maximum
likelihood robust (MLR) estimation method. To ascertain the
adequacy of model fit, we used the comparative fit index (CFI),
non-normed fit index (NNFI, also known as the Tucker-Lewis index),
root-mean-square error of approximation (RMSEA), and χ2 test
statistic. The NNFI and CFI usually vary along a 0- to-1 continuum
(the NNFI can be greater than 1 because of sampling, but this is
rarely the case in practice) in which values greater than 0.90 and
0.95 typically reflect acceptable and excellent fit to the data,
respectively (Schumacker & Lomax, 1996). Browne, Cudeck and
Bollen (1993) suggest that RMSEAs below 0.05 are indicative of a
“close fit” and that values of up to 0.08 represent reasonable
errors of approximation. Whereas the NNFI and RMSEA contain a
penalty for lack of parsimony, the CFI does not, such that the
addition of new parameters leads to an improved fit that may
reflect capitalization on chance. Using the Akaike Information
Criteria (AIC; Akaike, 1987) and Baysian Information Criteria (BIC)
also facilitates model comparison. In our study, model comparison
was guided by the following criteria: (a) a decrease of more than
0.01 in CFI and NNFI values when comparing a less restrictive model
to a more restrictive one is indicative of some model misfits; (b)
a decrease of 0.015 in RMSEA values when comparing a less
restrictive model to a more restrictive one is indicative of some
model misfits; and (c) lower values for the BIC and AIC are
indicative of better model fit.
Missing data ranged from less than 1% to 38% (see Table 1). As
in most longitudinal studies,
there were more missing values at T2. Little’s (1988) test was
computed for all individual indicators using the Missing Values
Analysis add-on module in SPSS. The p value was statistically
significant (χ2 [2229] = 2699.07, p = .001), indicating that the
data were not missing completely at random (MCAR). In addition,
because more data were missing at T2, we tested for differences in
various model parameters at T1 between completers and partial
completers (see Table 2). That is, those who did not participate at
T2 were compared to those who participated at both time points.
Overall, the invariance analyses showed no differences in the
various parameters (Models 1 to 6; see Table 2), except for
variances/covariances, where the CFI and NNFI values decreased by
0.01. However, no substantial latent mean differences in latent
constructs across the completers and partial completers were
observed, since none of the fit indices improved substantially when
these means were relaxed in Model 6. This indicates that the
results based on the full information
-
11
maximum likelihood (FIML) solution for estimating subsequent
models could be interpreted with a higher level of confidence. This
led us to conclude that, while the data were not MCAR, there were
no systematic patterns of missingness that could seriously bias
subsequent results based on CFA or structural equation modeling
(SEM). To account for missing data in the SEM analyses, FIML was
used to compute the product of individual likelihood functions in
order to estimate the analysis parameters. Using an FIML procedure
under MLR to treat missing data is considered superior to using
listwise deletion and other ad hoc methods, such as mean
substitution (Davey, Shanahan, & Schafer, 2001; Peugh &
Enders, 2004).
2.3.1. CFA models
To appropriately analyze the direction of effects or causal
ordering among variables, it is important to specify CFA and SEM
models that encompass the estimation of various statistical
parameters. First, all latent constructs should be measured at
least twice over a certain period of time. This period should make
it possible to capture a sufficient amount of change. In this
study, like many others in the field of educational psychology
(e.g., Guay, Marsh, & Boivin, 2003), we used a one-year
interval. Second, latent constructs should be measured using
multiple indicators. In this study, school attachment was measured
through five indicators, while other latent constructs were
assessed through four indicators. Third, uniquenesses (error) of
the same indicators (item) measured at two time points (T1 and T2)
should be correlated to prevent an inflated stability effect (a
method bias) of a latent construct measured at T1 and T2. In this
study, for all latent constructs measured at two time points,
uniquenesses at T1 and T2 were correlated. For example, the
uniqueness of item 1 of the intrinsic motivation latent construct
at T1 was correlated with the uniqueness of item 1 of the intrinsic
motivation latent construct at T2. Moreover, as mentioned above,
relatedness items were the same for each source (parents, friends
and teachers). When the items are the same, inflated correlations
(a method bias) might occur between relatedness latent constructs.
To avoid such a method effect, we estimated correlated uniquenesses
among items having the same wording (i.e., parallel items). For
instance, the uniqueness of item 1 for relatedness with friends’
latent construct was correlated with the uniquenesses of item 1 for
relatedness with teachers and parents. According to Podsakoff,
MacKenzie, Lee, and Podsakoff (2003), correlated uniqueness models
are quite flexible and likely to converge and produce proper
parameter estimates. However, they are characterized by some
weaknesses, including that method effects are orthogonal, the trait
variance is possibly biased, and the trait and method effects do
not interact. Because the models tested in our research involved
many free parameters, this correlated uniqueness model – being much
more flexible than other models used to control for potential bias
(i.e., multiple-method-factor-approach) – appeared the most
appropriate. In sum, considering it to be the best statistical
approach, we chose the correlated uniquenesses model to control for
method bias in our research. Fourth, it is important to make sure
that students have the same interpretation
-
12
of the latent construct over time. As a result, a model was
estimated with an equality constraint on the factor loadings over
time. We tested two CFA models: one containing all the statistical
parameters described above, including all covariances among the
twelve latent constructs, and another with the exact same
parameters, but with factor loadings that were invariant through
time.
2.3.2. SEM models The SEM models included all the statistical
parameters described for the CFA models except
that latent covariances were remodeled in accordance with the
hypothesized model in which correlations among independent
variables at T1 and disturbances at T2 were estimated. The
correlation among independent variables allows the estimation of
hypothesized causal effects while controlling for stability
effects. Because it is highly unlikely that independent variables
will explain all the variance in the dependent variables,
correlations among disturbance terms are estimated among all
dependent variables. We tested four SEM models, corresponding to
Models 3 to 6 in Table 3: Model 3, containing only stability paths;
Model 4, containing stability paths and the effect of all sources
of relatedness and school attachment on intrinsic and identified
regulation; Model 5, containing reciprocal effects between sources
of relatedness and school attachment, and the two types of
regulation; and Model 6, containing reciprocal effects and
inter-dependency effects among sources of relatedness and school
attachment. Modest effect sizes for the cross-lagged paths were
expected because the latent constructs were likely to be relatively
stable over time. For this reason, if the NNFI containing a penalty
for lack of parsimony did not decrease compared to the other more
parsimonious models (e.g., stability and unidirectional), the more
global model would be interpreted, especially if some significant
effects were observed.
2.3.3. Gender invariance analyses
To test for possible moderating effects of gender, we built a
series of five invariance models where equality constraints were
imposed on various statistical parameters (Models 7 to 11; see
Table 3). In the least restrictive model, factor loadings and
intercepts were fixed to be equal across boys and girls. These
equality constraints ensure that the results can be meaningfully
compared across groups. In subsequent models, uniquenesses,
correlated uniquenesses, variances and covariances, and path
coefficients were constrained to equality (see Guay, Morin,
Litalien, Valois, & Vallerand, 2015). Finally, we explored
gender mean differences in all latent constructs by fixing
-
13
means to 0 in one model in order to compare it to another model
where means were relaxed for girls.
3. Results
3.1. CFA
Descriptive statistics for all items used in this study are
presented in Table 1, and fit indices
for the various tested models are presented in Table 3. To
provide an overview of the correlation pattern, we conducted a CFA
in which no constraint on the various parameters were imposed. This
model estimated all correlations among the uniquenesses described
above. Standardized factor loadings and uniquenesses are presented
in Table 1, model fit indices are presented in Table 3 (Model 1),
and correlations among the six factors at T1 and T2 are presented
in Table 4. First, all standardized factor loadings were above
0.40. Second, cross-sectional correlations among the sources of
relatedness and school attachment at T1 and T2 were low enough to
indicate that the constructs, while related, appeared to be
distinct. Neither the correlations among the sources of relatedness
nor those connecting these sources to school attachment exceeded
0.50. Moreover, when we used the longitudinal data measurement
points in this study as a different method in order to build a
multitrait-multimethod (MTMM) correlation matrix (Campbell &
Fiske, 1959), we noted that convergent correlations (i.e., between
two different methods, namely T1 and T2, assessing the same latent
constructs) were higher than divergent ones (i.e., between two
different latent constructs assessed using different methods). More
specifically, the longitudinal correlations connecting various
sources of relatedness or school attachment over time (divergent
correlations) were weaker than the longitudinal relationships
between the same constructs (relatedness or school attachment) over
time. This first pattern of results is especially important because
sources of relatedness or school attachment that were too strongly
correlated might have compromised the first goal of this study,
which was to determine which variables were most likely to predict
intrinsic and identified regulation.
Third, if we look at correlations connecting both types of
regulation (intrinsic and identified regulation) to sources of
relatedness and school attachment, the results indicate that
intrinsic and identified regulation were most strongly correlated
with school attachment, followed in decreasing order by relatedness
with teachers, parents and friends, respectively. This pattern of
cross-sectional correlations was similar at both time points.
Moreover, this pattern of results was reproduced longitudinally.
For example, school attachment at T1, as compared to relatedness
with friends at T1, was more strongly correlated with T2 intrinsic
and identified regulation.
-
14
A subsequent CFA model was estimated to determine whether factor
loadings were invariant over time (see Model 2 in Table 3). Given
that the NNFI and CFI increased, the RMSEA was identical to Model
1, and the BIC and AIC values were lower, we considered the factor
loadings to be invariant across time. These results mean that, over
time, students had the same understanding of items designed to
measure a construct. Thus, the absence of longitudinal
relationships could not be explained by the fact that students
understood the construct differently over time. Based on these
results, we estimated SEM models with these time-in-variant
constraints.
3.2. SEM
As mentioned above, four SEM models were tested: horizontal
(i.e., stability over time of a given latent variable),
unidirectional (i.e., sources of relatedness and school attachment
predicting types of regulation), reciprocal (i.e., both directions
among sources of relatedness/ school attachment and types of
regulation), and reciprocal and inter-dependency effects (see Table
3, Models 3–6). The fit indices for these four models were nearly
equivalent, except for Model 6, where we observed a slight decrease
(0.001) in the NNFI value. Inspection of Model 6 revealed no
significant interdependencies, meaning that different sources of
relatedness or school attachment were not significantly related
over time. Because the NNFI of Model 5, which contains a penalty
for lack of parsimony, remained unchanged by the addition of
cross-lagged paths, we determined that this was the best model and
based our main interpretations on it. Results of Model 5 are
presented in Table 5. Inspections of path coefficients revealed
that stability effects for all latent constructs were significant
and ranged between 0.39 and 0.60. One cross-lagged path was
statistically significant at p < .05, namely that connecting T1
school attachment to T2 intrinsic motivation (b = 0.12), and two
cross-lagged paths were statistically significant at p < .10,
namely that connecting T1 relatedness with teachers to T2
identified regulation (b = 0.11) and that connecting T1 relatedness
with parents to T2 intrinsic motivation (b = −0.08).
-
15
This last finding should be interpreted with caution, especially
since this relationship was positive in the correlation matrix (see
Table 4 for correlation coefficients). These results suggest that
relatedness with parents acted as a suppressor variable in the
statistical model. Suppression is a statistical phenomenon that can
occur in regression analyses when predictors are strongly
correlated. In such cases, the relationship between two variables
will be artificially inverted (Cohen & Cohen, 1983). According
to Maassen and Bakker (2001), suppression can occur in structural
equation modeling and the probability of occurrence is relatively
high in models that have latent variables. The fact that perceived
relatedness with parents at T1 was positively correlated with
intrinsic motivation at T2 at the bivariate level suggests that the
negative path was probably a spurious one. In addition, three paths
connecting regulation types to sources of relatedness/school
attachment were marginally significant (p < .10; see Table 5),
namely intrinsic regulation to relatedness with friends (b = 0.10),
intrinsic regulation to school attachment (b = 0.10), and
identified regulation to school attachment (b = 0.10). However,
caution should be exercised when interpreting these findings
because their magnitude was low (approx. 0.10) and some of them
were significant at p < .10. As discussed in the next section,
invariance analyses were conducted as a function of gender based on
the reciprocal effect model (Model 5; see Table 3).
3.3. Gender
invariance analyses
The five models (Models 7 to 11; see Table 3) indicated that
factor loadings, intercepts,
uniquenesses, correlated uniquenesses and path coefficients were
invariant across gender. More specifically, for Models 7 to 9,
equality constraints imposed on various statistical parameters did
not result in lower CFI, NNFI or RMSEA values or in higher BIC or
AIC values. These results ensure that the various statistical
parameters can be meaningfully compared across groups because they
are not due to a different understanding of the items for boys and
girls. However, imposing equality constraints on
variances/covariances resulted in lower fit indices or higher AIC
or BIC values. For this reason, we specified Model 12, in which
variances/covariances were relaxed. Model 12 had similar fit
indices to Model 7, the least restrictive model tested for
invariance analyses. We thus determined that Model 12 was the best
model and based our interpretations on it. Results of this model
are depicted in Fig. 1 and Table 5. All coefficients depicted were
standardized. First, stability paths were relatively high, ranging
from 0.38 to 0.61. Second, school attachment at T1 was positively
and significantly (p < .05) associated with intrinsic motivation
at T2 (b = 0.11 for boys, b = 0.12 for girls). Contrary to what we
observed in the general model, relatedness with teachers at T1 was
now positively and significantly associated with identified
regulation at T2 (b = 0.10 for boys, b = 0.13 for girls). No other
cross-lagged paths were significant. These findings indicate that
the results concerning the path coefficients were the same across
gender. The only differences we found concerned variances and
covariances. However, because these parameters are less relevant in
this study, they were not interpreted.
Based on these results, we tested two subsequent CFA models
(Model 13 and 14; see Table
3) to determine whether there were gender mean differences in
the 12 latent constructs. Because the NNFI decreased slightly when
means were set to 0 compared to a model where these gender mean
differences were estimated, we decided to interpret mean
differences across gender based on Model 13. When boys’ latent
means at T1 were fixed to 0 for identification purposes, girls’
latent means (expressed as differences in standard deviation units)
were significantly higher for T1 intrinsic motivation (M = 0.15; SE
= 0.08; p < .05) and for T1 and T2 identified regulation (T1:
M=0.21; SE=0.09; p < .05; T2: M=0.28; SE=0.12; p < .05), but
lower for T1 and T2 relatedness
-
16
with parents (T1: M = -19; SE=0.06; p < .05; T2: M=−0.29;
SE=0.07, p < .05). There were no significant differences for
school attachment or relatedness with friends.
4. Discussion
This study pursued four goals: (a) to test whether multiple
sources of relatedness and school attachment predicted intrinsic
and identified regulation over an one-year period; (b) to determine
whether there were interdependency effects among sources of
relatedness and school attachment; (c) to investigate whether there
were reciprocal relationships among sources of relatedness, school
attachment, and the two regulation types; and (d) to test whether
gender moderated the relationships connecting sources of
relatedness and school attachment to intrinsic and identified
regulation. Findings based on several structural equation models
indicate that, in our sample of adolescent students from
disadvantaged neighborhoods, only relatedness with teachers and
school attachment positively predicted intrinsic and identified
regulation. More specifically, relatedness with teachers predicted
identified regulation, whereas school attachment was positively
linked to intrinsic regulation. In addition, no interdependency
effects were observed among the sources of relatedness and school
attachment, and neither regulation type was found to have
significant effects on the sources of relatedness or school
attachment. Finally, gender did not moderate the observed
associations.
Although our hypothesis based on the additive model was not
supported, our findings still
extend previous studies on sources of relatedness, school
attachment and adolescents’ school-related functioning in several
important ways. Indeed, our results provide evidence that school
attachment and relatedness with teachers are the only two relevant
predictors of intrinsic and identified regulation for school-based
learning activities. Although previous research and theory have
underscored the importance of relatedness with teachers in
predicting autonomous regulation (Ryan et al., 1994), this study
underscores the role of school attachment in intrinsic regulation,
which has received less attention in previous research. In line
with these ideas, our results also
-
17
highlight that, for these regulation outcomes, different sources
of relatedness play different roles, a finding that needs further
consideration. It is reasonable to expect that teachers who spend a
lot of time with students will create a classroom context that is
more favorable to the internalization of the value of school
activities (i.e., identified regulation). However, it is surprising
that teachers did not play a significant role in promoting
intrinsic motivation. School attachment might be a salient global
feeling that helps students negotiate various school activities
effectively by keeping their levels of intrinsic motivation
relatively high. This feeling of attachment is developed because
the school, as an independent entity per se, is able to meet
students’ needs for relatedness through numerous activities (e.g.,
cultural and sports activities). Nevertheless, the magnitude of the
effects of school attachment and relatedness with teachers was
small, based on Cohen’s (1992) guidelines on effect sizes (10 =
small, 0.30 = medium, 0.50 = large). Yet, our findings are
particularly notable in that they were based on a cross-lagged
autoregressive model controlling for prior levels of the dependent
variable, whereas most other studies have used cross-sectional data
or partial longitudinal models. The stability effects in
autoregressive models are usually high, although they can be
attenuated by correlated uniquenesses. These stability coefficients
offer the possibility of testing more robust effects of one latent
construct on a subsequent latent construct. However, they have an
impact on the magnitude of the cross-lagged coefficients. It should
be noted that a small coefficient (e.g., 0.12) obtained over a
one-year period might have a cumulative impact on the outcome over
several years of schooling. Although there are no clear guidelines
for interpreting these autoregressive coefficients, we believe that
they should be analyzed carefully in light of the stability effects
and the various predictors considered simultaneously in the
analyses.
It is noteworthy that neither relatedness with friends nor
relatedness with parents significantly predicted the two types of
regulation. Moreover, these findings do not appear to be
attributable to inter-dependency effects or a degree of
transference. Indeed, cross-lagged analyses revealed no effect of
prior relatedness with friends or parents on school attachment or
relatedness with teachers. This means that the lack of any
significant predictive associations between relatedness with
parents and friends and the two types of regulation were not
explained by mediation processes through school attachment or
relatedness with teachers. A certain degree of transference between
friends and school attachment, as well as between parents and
teachers, might be expected. Those who feel attached to their
friends may experience greater feelings of school attachment, and
those who have positive and trust-worthy relationships with their
parents may experience greater relatedness with their teachers. The
fact that such transference did not appear to occur increases our
level of confidence regarding the lack of effect of relatedness
with friends and parents on the two types of regulation. Such
non-significant findings are nevertheless surprising given that
many studies in the field have outlined the fundamental role of
parents and friends/peers with respect to constructs akin to
identified and intrinsic regulation for school activities (e.g.,
Vallerand et al., 1997; Wentzel, Russell, & Baker, 2015). One
reason may be that the present study encompasses various sources of
relatedness and school attachment, which has not always been the
case in previous studies (e.g., Guay et al., 2008). Because most
sources of relatedness and school attachment were positively
correlated, controlling for them in a longitudinal design provided
a stringent test of the threshold or additive model. More
specifically, it appears that above a certain degree of school
attachment and relatedness with teachers, relatedness with friends
and relatedness with parents were redundant, such that these
additional sources of relatedness did not improve either type of
regulation. Consequently, these results might challenge the
sensitization process proposed by Moller et al. (2010), who showed
that relatedness leads to an incremental value of relatedness for
intrinsic and identified regulation. According to this perspective,
the more people experience positive relatedness in various
relationships, the more they benefit from them. In contrast,
our
-
18
results are more in line with the satiety perspective. However,
no definitive conclusion can yet be reached because the method used
in our study was starkly different from that used by Moller et al.
(2010). On the other hand, our results are in line with other
studies in the field of achievement (Hattie, 2009) showing that
teachers are the most significant source of influence on students’
achievement. While the influence of parents and peers is relevant,
it is less important than that of teachers and school
attachment.
To our knowledge, only Guay et al. (2008) have compared the
effects of friends and parents
on different types of regulation in auto-regressive models
involving a sample of young adults at college (18 years old). Their
results indicate that only relatedness with parents predicted
subsequent levels of intrinsic and identified regulation.
Nevertheless, since their study did not include relatedness with
teachers or attachment to college, it is not known whether the
results for relatedness with parents would be corroborated if these
variables were taken into account. However, it is important to note
that different sources of relatedness might play different roles
depending on the developmental period, as proposed by some theories
(De Goede et al., 2008; Harter, 1999). Relatedness with parents
might be more relevant during the elementary school years than
during adolescence, and the importance of parents might
subsequently be restored when adolescents develop into young adults
facing important life decisions. Further studies taking a
developmental approach are thus needed to answer this question.
Of additional interest is the fact that there were no effects of
prior types of regulation on
subsequent sources of relatedness (i.e., no reciprocal effects),
thereby providing support for our initial hypothesis based on SDT.
These findings indicate that students who experienced pleasure,
satisfaction and enjoyment as well as those who had internalized
the importance of school did not report greater feelings of
relatedness in various relationships or greater school attachment.
According to Guay et al. (2008), such a lack of significant results
could stem from the fact that, during adolescence, sources of
relatedness and school attachment are relatively stable and thereby
attenuate the possibility of various regulation types predicting
these variables. However, the regulation types in our study
appeared to be much more stable over time than the different
sources of relatedness and school attachment, thereby providing
weak support for such an interpretation. A simpler interpretation
of these findings may be that the direction of the effects was from
sources of relatedness and school attachment to various regulation
types, which is in line with SDT (Ryan & Deci, 2009).
As expected, we found no evidence for the moderating effect of
gender. Not only were boys
and girls both affected by relatedness with teachers and school
attachment, but they were also affected to a very similar degree
(i.e., path coefficients were not significantly different). Such
results are consistent with previous findings showing that
relationships between sources of relatedness and various regulation
types are usually invariant (Guay et al., 2003; Rueger et al.,
2010; Wang & Eccles, 2012). Based on these results, we conclude
that the function served by these sources is much more similar than
different across gender, as suggested by Burleson (2003). Moreover,
such findings are in line with SDT (Ryan & Deci, 2009), which
posits that the need for relatedness serves the same function
across various groups of individuals, including males and
females.
However, it is still important to keep in mind that gender
differences existed at the mean level,
although such differences were not a primary focus in this
investigation. More specifically, compared to boys, girls’ latent
means were significantly higher for T1 intrinsic motivation and
T1
-
19
and T2 identified regulation, but lower for T1 and T2
relatedness with parents. There were no significant differences for
school attachment or relatedness with friends and teachers,
contrary to our initial hypothesis. The results regarding intrinsic
and identified regulation were consistent with past studies showing
that girls are more intrinsically motivated toward school and
internalize the value of schooling more than boys (Vallerand et
al., 1997). However, the gender differences pertaining to
relatedness with parents are more surprising. How to explain the
fact that the girls in our sample reported lower levels of
relatedness with parents than the boys? Some previous studies have
shown that girls experience more support from their parents than
boys (Guay et al., 2003), whereas other studies have reported no
significant gender differences (Rueger et al., 2010). Our sample
characteristics might explain why our findings differ from past
studies. More specifically, students in this sample attended
schools in socio-economically disadvantaged neighborhoods.
Consequently, it is possible that, in at-risk family contexts,
girls experience more conflicts with their parents than boys, which
may have led to lower scores on the relatedness measure. This
tentative explanation should, however, be further examined in
future studies. Such gender differences might, however, lead to
gender-specific intervention strategies. For instance, parents of
girls attenting a socioeconomically disadvantaged school might be
informed that the quality of the relationship they have with their
daughter is important, even if the associations connecting
relatedness with parents to both types of regulations were not
statistically significant in this study. Indeed, previous research
has shown that positive parenting behaviors predict other important
outcomes such as emotional adjustment (Ratelle, Duchesne, &
Guay, 2017).
Again, it is important to note that the results were obtained
among a sample of students
attending schools in disadvantaged neighborhoods. It is possible
that many youth attending such schools do not view school as a
relevant pathway to achieving their future goals, and thus feelings
of warmth toward teachers (or attachment to school) may not be as
strongly associated with academic motivation as for students
attending schools in more affluent neighborhoods.
4.1. Limitations and conclusions The findings from this study
should be interpreted in light of certain limitations. First,
although we used an auto-regressive longitudinal design, it is
important to keep in mind that the meaning of the term “effects”
remains tentative. The correlational nature of the data precludes
any firm conclusion regarding the direction of causality among the
constructs. Second, we asked students to self-report on various
sources of relatedness and school attachment. Such self-reported
measures are subject to shared method variance as well as potential
bias in the estimation of scores. More specifically, this
single-informant approach might have led to an inflated
relationship between the sources of relatedness and school
attachment and the two types of regulation. In other words, the
reciprocal relationships could also be mere perceptions – when
students enjoy school or value it, they are more likely to perceive
their social context more positively. Third, the observed effect
sizes were relatively small. However, the magnitude of the
cross-lagged effects observed in this study was consistent with
previous studies (e.g., Guay et al., 2008). Fourth, there were a
large number of missing values in this study. Less than one third
of contacted youth returned surveys at Time 1. It is possible that
the two thirds of youth who chose not to participate were less
attached to school, had lower school enjoyment, and valued school
less than the students who participated. Yet, the results from the
invariance analyses and SEM model indicated few differences between
the participants who provided complete data and those for whom
values were missing. In addition, we used a sophisticated approach
to deal with missing values (FIML). Therefore, we believe that
missing data did not compromise the validity of our study.
Nevertheless, it is quite possible that
-
20
our sample is not representative of the entire high school
population from disadvantaged neighborhoods, since our students may
have been more academically oriented and well adjusted than most
students in this population. Finally, we only measured intrinsic
and identified regulation and these motivational measures were not
specific to school subjects. It is thus important to exercise
caution in generalizing these results.
Further research should test the replicability of our results in
different populations, including
students from different cultural and socioeconomic backgrounds.
In addition, because the regression coefficients are low, perhaps
there are theoretically plausible moderator or mediator variables
that could be tested in future research. For example, intrinsic and
identified regulations of students scoring low on neuroticism might
by more positively affected by relationships with teachers, parents
and peers than those high in neuroticism. Furthermore, some
qualitative research may be in order to examine the experience of
school attachment and sources of relatedness. In addition, it could
be quite useful to study not only positive characteristics of
parents, teachers and friends, but also negative characteristics,
including abusive control from parents and teachers and negative
peer experiences such as victimization and rejection. Indeed, some
students may experience negative social experiences that could
demotivate them at school.
Finally, it is important to emphasize the practical implications
of our results. First, it may be
important to inform schools that they have a role to play in
adolescents’ perceptions of school attachment and, subsequently, in
their intrinsic motivation. To this end, schools could provide
opportunities for and reward positive involvement, and enable
students to develop the skills needed to take advantage of these
opportunities. For example, programs intended to reduce school
bullying (Schneider, O'Donnell, Stueve, & Coulter, 2012) or
promote school-based extracurricular activities (Denault &
Guay, 2017) might help achieve this goal. Moreover, teachers should
be informed that the sense of relatedness they develop with their
students is an important catalyzer of how students endorse the
importance of school-related activities. In this regard, many
pedagogical practices might be used to foster a positive sense of
relatedness, including autonomy support (offering choices),
involvement (being interested in the students experiences), and
structure (providing feedback; Ryan & Deci, 2009).
In sum, this longitudinal research has moved the field forward
in some meaningful ways,
theoretically, conceptually, empirically and practically.
Theoretically, this study provides somewhat of a field test of the
sensitization perspective proposed by SDT and shows that it is the
social context that leads to inter-individual changes in motivation
type, rather than the reverse. Conceptually, it also offers the
possibility to delineate sources of relatedness from school
attachment. Empirically, the statistical models and the method used
reflect state of the art techniques in the field of applied
statistics. Finally, the practical implications outlined are
important for school administrators and teachers.
Appendix A. Supplementary data Supplementary data associated
with this article can be found, in the online version, at
http://dx.doi.org/10.1016/j.cedpsych.2017.10.001
-
21
References
Akaike, H. (1987). Factor analysis and AIC. Psychometrika,
52(3), 317–332. Anderman, L. H. (2003). Academic and social
perceptions as predictors of change in middle school
students' sense of school belonging. The Journal of Experimental
Education, 72(1), 5–22.
Barber, B. K., & Olsen, J. A. (1997).
Socialization in context connection, regulation, and autonomy
in the family, school, and neighborhood, and with peers. Journal
of Adolescent Research, 12(2), 287–315.
Battistich, V., Solomon, D., Kim, D., & Watson, M. (1995).
Schools as communities, poverty
levels of student populations, and students' attitudes, motives,
and performance: A multilevel analysis. American Educational
Research Journal, 32, 627–658.
Baumeister, R. F., & Leary, M. R. (1995). The need to
belong: desire for interpersonal attachments
as a fundamental human motivation. Psychological Bulletin, 117,
497–529.
Browne, M. W., Cudeck, R., & Bollen, K. A. (1993).
Alternative ways of assessing model fit. Sage
Focus Editions, 154.
http://dx.doi.org/10.1177/0049124192021002005 136-136. Burleson, B.
R. (2003). The experience and effects of emotional support: What
the study of cultural
and gender differences can tell us about close relationships,
emotion, and interpersonal communication. Personal Relationships,
10(1), 1–23.
Burney, V. H., & Beilke, J. R. (2008). The constraints of
poverty on high achievement. Journal
for the Education of the Gifted, 31, 171–197. Burton, K. D.,
Lydon, J. E., D’Alessandro, D. U., & Koestner, R. (2006). The
differential effects
of intrinsic and identified motivation on well-being and
performance: Prospective, experimental, and implicit approaches to
self-determination theory. Journal of Personality and Social
Psychology, 91, 750–762.
Campbell, D. T., & Fiske, D. W. (1959). Convergent and
discriminant validation by the multitrait-
multimethod matrix. Psychological bulletin, 56(2), 81. Chiu, M.
M., Chow, B. W. Y., McBride, C., & Mol, S. T. (2016). Students’
sense of be- longing
at school in 41 countries: Cross-cultural variability. Journal
of Cross-Cultural Psychology, 47(2), 175–196.
Cohen, J., & Cohen, P. (1983). Applied multiple
regression/correlation analysis for the behavioral
sciences. Hillsdale, NJ: L. Erlbaum. Cooper, H., Lindsay, J. J.,
& Nye, B. (2000). Homework in the home: How student, family,
and
parenting-style differences relate to the homework process.
Contemporary Educational Psychology, 25, 464–487.
-
22
Crosnoe, R., Johnson, M., & Elder, G. (2004).
Intergenerational bonding in school: The behavioral and contextual
correlates of student–teacher relationships. Sociology of
Education, 77, 60–81.
Davey, A., Shanahan, M. J., & Schafer, J. L. (2001).
Correcting for selective nonresponse in the
National Longitudinal Survey of Youth using multiple imputation.
Journal of Human Resources, 500–519.
Davidson, A. J., Gest, S. D., & Welsh, J. A. (2010).
Relatedness with teachers and peers during
early adolescence: An integrated variable-oriented and
person-oriented approach. Journal of School Psychology, 48(6),
483–510.
De Goede, I. H., Branje, S. J., & Meeus, W. H. (2008).
Developmental changes in adolescents’ perceptions of relationships
with their parents. Journal of Youth and Adolescence, 38(1), 75–88.
Deci, E. L., & Ryan, R. M. (2000). The “what” and “why” of goal
pursuits: Human needs and the
self-determination of behavior. Psychological inquiry, 11(4),
227–268. Deci, E. L., & Ryan, R. M. (1991). A motivational
approach to self: Integration in personality. In
E. L. Deci, & R. M. Ryan (Eds.). Perspectives on motivation.
Nebraska symposium on motivation (pp. 237–288). Lincoln, NE:
University of Nebraska Press.
Denault, A. S., & Guay, F. (2017). Motivation towards
extracurricular activities and motivation at
school: A test of the generalization effect hypothesis. Journal
of Adolescence, 54, 94–103. Donnelly, L. (2015). Neighborhood
disadvantage and school dropout: A multilevel analysis of
mediating contexts. Doctoral dissertationNew Brunswick: Rutgers
University-Graduate School.
Eccles, J. S., Midgley, C., Wigfield, A., Buchanan, C. M.,
Reuman, D., Flanagan, C., et al. (1993).
Development during adolescence: The impact of stage-environment
fit on young adolescents’ experiences in schools and in families.
American Psychologist, 48, 90–101.
Fantuzzo, J., Stoltzfus, J., Lutz, M. N., Hamlet, H., Balraj,
V., Turner, C., et al. (1999). An
evaluation of the special needs referral process for low-income
preschool children with emotional and behavioral problems. Early
Childhood Research Quarterly, 14, 465–482.
Furrer, C., & Skinner, E. (2003). Sense of relatedness as a
factor in children's academic
engagement and performance. Journal of Educational Psychology,
95(1), 148. Gilligan, C. (1982). In a different voice. Harvard
University Press.
Goodenow, C. (1993). The
psychological sense of school membership among adolescents:
Scale development and educational correlates. Psychology in the
Schools, 30, 79–90.
Goodenow, C., & Grady, K. E. (1993). The relationship of
school belonging and friends' values to
academic motivation among urban adolescent students. The Journal
of Experimental Education, 62(1), 60–71.
-
23
Guay, F., Marsh, H. W., & Boivin, M. (2003). Academic
self-concept and academic achievement:
Developmental perspectives on their causal ordering. Journal of
Educational Psychology, 95(1), 124.
Guay, F., Marsh, H. W., Senécal, C., & Dowson, M. (2008).
Representations of relatedness with
parents and friends and autonomous academic motivation during
the late adolescence–early adulthood period: Reciprocal or
unidirectional effects? British Journal of Educational Psychology,
78(4), 621–637.
Guay, F., Morin, A. J., Litalien, D., Valois, P., &
Vallerand, R. J. (2015). Application of
exploratory structural equation modeling to evaluate the
academic motivation scale. The Journal of Experimental Education,
83(1), 51–82.
Harter, S. (1999). The construction of the self: A developmental
perspective. New York: The
Guilford Press. Hattie, J. (2009). Visible learning: A synthesis
of over 800 meta-analyses relating to achievement.
NY: Routledge. Hawkins, J. D., Guo, J., Hill, K. G.,
Battin-Pearson, S., & Abbott, R. D. (2001). Long-term
effects
of the Seattle Social Development Intervention on school bonding
trajectories. Applied Developmental Science, 5, 225–236.
Hill, J. P., & Lynch, M. E. (1983). The intensification of
gender-related role expectations during
early adolescence. In J. Brooks-Gunn, & A. C. Peterson
(Eds.). Girls at puberty (pp. 201–228). Springer.
Hill, L. G., & Werner, N. E. (2006). Affiliative motivation,
school attachment, and ag- gression in
school. Psychology in the Schools, 43(2), 231–246. Huebner, E.
S., Gilman, R. I. C. H., & Furlong, M. J. (2009). A conceptual
model for re- search in
positive psychology in children and youth. In Handbook of
positive psychology in schools (pp. 3–8).
In’nami, Y., & Koizumi, R. (2013). Structural equation
modeling in educational research: A
primer. In M. Swe Khine (Ed.). Application of structural
equation modeling in educational research and practice (pp. 23–54).
Rotterdam: Sense Publishers.
Laursen, B., & Mooney, K. S. (2008). Relationship network
quality: Adolescent adjustment and
perceptions of relationships with parents and friends. American
Journal of Orthopsychiatry, 78(1), 47.
Maassen, G. H., & Bakker, A. B. (2001). Suppressor variables
in path models: Definitions and
interpretations. Sociological Methods & Research, 30(2),
241–270.
-
24
Maccoby, E. E. (1998). The two sexes: Growing up apart, coming
together. Harvard University Press.
Ministère de l’Éducation, du Loisir et du Sport (2007–2008).
Indice de défavorisation par école.
Québec, QC: Gouvernement du Québec. Moller, A. C., Deci, E. L.,
& Elliot, A. J. (2010). Person-level relatedness and the incre-
mental
value of relating. Personality and Social Psychology Bulletin,
36, 754–767. Muthén, L. K., & Muthén, B. O. (2012). Mplus
user’s guide (7th ed). Los Angeles, CA: Muthén
& Muthén.
Niehaus, K., Rudasill, Moritz, & Rakes, C. R.
(2012). A longitudinal study of school
connectedness and academic outcomes across sixth grade. Journal
of School Psychology, 50, 443–460.
O’Brennan, L., Pas, E., & Bradshaw, C. (2017). Multilevel
examination of burnout among high
school staff: Importance of staff and school factors. School
Psychology Review, 46(2), 165–176.
Oyserman, D. (2013). Not just any path: Implications of
identity-based motivation for disparities
in school outcomes. Economics of Education Review, 33, 179–190.
Peterson, C. (2006). A primer in positive psychology. Oxford
University Press.
Peugh, J. L., & Enders, C. K. (2004).
Missing data in educational research: A review of reporting
practices and suggestions for improvement. Review of Educational
Research, 74(4), 525–556.
Podsakoff, P. M., MacKenzie, S. B., Lee,
J. Y., & Podsakoff, N. P. (2003). Common method biases
in behavioral research: A critical review of the literature and
recommended remedies. Journal of Applied Psychology, 88(5),
879.
Ratelle, C. F., Duchesne, S., & Guay, F. (2017). Predicting
students’ adjustment from multiple
perspectives on parental behaviors. Journal of Adolescence, 54,
60–72. Ratelle, C. F., Guay, F., Vallerand, R. J., Larose, S.,
& Senécal, C. (2007). Autonomous,
controlled, and amotivated types of academic motivation: A
person-oriented analysis. Journal of Educational Psychology, 99(4),
734.
Richards, J., Wiese, C., Katon, W., Rockhill, C., McCarty, C.,
Grossman, D., et al. (2010).
Surveying adolescents enrolled in a regional healthcare delivery
organization: Mail and phone follow-up—what works at what cost?
Journal of the American Board of Family Medicine, 23(4), 534–541.
http://dx.doi.org/10.3122/jabfm.2010.04.100019.
Rudasill, K. M., Niehaus, K., Crockett, L. J., & Rakes, C.
R. (2014). Changes in school
connectedness and deviant peer affiliation among sixth-grade
students from high- poverty
-
25
neighborhoods. The Journal of Early Adolescence, 34, 896–922.
Rueger, S. Y., Malecki, C. K., & Demaray, M. K. (2010).
Relationship between multiple sources
of perceived social support and psychological and academic
adjustment in early adolescence: Comparisons across gender. Journal
of Youth and Adolescence, 39(1), 47–61.
Ryan, R. M. (1993). Agency and organization: Intrinsic
motivation, autonomy and the self in
psychological development. In J. Jacobs (Vol. Ed.), Nebraska
Symposium on Motivation: Developmental perspectives on motivation:
Vol. 40, (pp. 1–56). Lincoln, NE: University of Nebraska Press.
Ryan, R. M., & Deci, E. L. (2009). Promoting self-determined
school engagement. In K. R.
Wentzel, & A. Wigfield (Eds.). Handbook of motivation at
school (pp. 171–195). New York, NY: Routledge.
Ryan, R. M., Stiller, J. D., & Lynch, J. H. (1994).
Representations of relationships to teachers,
parents, and friends as predictors of academic motivation and
self-esteem. The Journal of Early Adolescence, 14(2), 226–249.
Schneider, S. K., O'Donnell, L., Stueve, A., & Coulter, R.
W. (2012). Cyberbullying, school
bullying, and psychological distress: A regional census of high
school students. American Journal of Public Health, 102(1),
171–177.
Schoon, I. (2008). A Transgenerational model of status
attainment: the potential med- iating role
of school motivation and education. National institute of
economic review, 205, 72–82. Scholte, R. H., & van Aken, M.
(2006). Peer relations in adolescence. In S. Jackson, & L.
Goossens
(Eds.). Handbook of adolescent development (pp. 175–199). New
York, NY, US: Psychology Press.
Schumacker, R. E., & Lomax, R. G. (1996). A guide to
structural equations modeling. Hillsdale,
NJ: Erlbaum. Senécal, C. B., Vallerand, R. J., & Vallières,
É. F. (1992). Construction et validation de l'Échelle
de la Qualité des Relations Interpersonnelles (EQRI). Revue
européenne de psychologie appliquée, 42, 315–322.
Vallerand, R. J., Blais, M. R., Briere, N. M., & Pelletier,
L. G. (1989). Construction et validation
de l'Échelle de motivation en éducation (EME). Canadian Journal
of Behavioural Science, 21, 323–349.
Vallerand, R. J., Fortier, M. S., & Guay, F. (1997).
Self-determination and persistence in a real-
life setting: Toward a motivational model of high school
dropout. Journal of Personality and Social Psychology, 72(5),
1161.
Vallerand, R. J., Pelletier, L. G., Blais, M. R., Briere, N. M.,
Sénécal, C., & Vallières, E. F. (1993).
On the assessment of intrinsic, extrinsic, and amotivation in
education: Evidence on the
-
26
concurrent and construct validity of the Academic Motivation
Scale. Educational and Psychological Measurement, 53, 159–172.
Wang, M. T., & Eccles, J. S. (2012). Social support matters:
Longitudinal effects of social support
on three dimensions of school engagement from middle to high
school. Child Development, 83(3), 877–895.
Wentzel, K. R., Russell, S., & Baker, S. (2015). Emotional
support and expectations from parents,
teachers, and peers predict adolescent competence at school.
Journal of Educational Psychology, 108(2), 242–255.
Youniss, J., & Smollar, J. (1985). Adolescent relations with
mothers, fathers, and friends. Chicago:
University of Chicago Press.