International Journal of School & Educational Psychology

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/usep20

Development of social skills during middle

childhood: Growth trajectories and school-related
predictors

Mari-Anne Sørlie, Kristine Amlund Hagen & Kristin Berg Nordahl

To cite this article: Mari-Anne Sørlie, Kristine Amlund Hagen & Kristin Berg Nordahl (2021)
Development of social skills during middle childhood: Growth trajectories and school-related
predictors, International Journal of School & Educational Psychology, 9:sup1, S69-S87, DOI:
10.1080/21683603.2020.1744492

To link to this article: https://doi.org/10.1080/21683603.2020.1744492

© 2020 The Author(s). Published with Published online: 12 May 2020.

license by Taylor & Francis Group, LLC.

Submit your article to this journal Article views: 25446

View related articles View Crossmark data

Citing articles: 9 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=usep20
INTERNATIONAL JOURNAL OF SCHOOL & EDUCATIONAL PSYCHOLOGY
2021, VOL. 9, NO. S1, S69–S87
https://doi.org/10.1080/21683603.2020.1744492

Development of social skills during middle childhood: Growth trajectories and

school-related predictors
Mari-Anne Sørlie, Kristine Amlund Hagen, and Kristin Berg Nordahl
Norwegian Research Centre AS (NORCE), The Norwegian Center for Child Behavioral Development (NUBU), Oslo, Norway

ABSTRACT KEYWORDS
This study examined the development of social skills across five measurement points from 4th Elementary school; growth
through 7th grade, and the influence of child gender and school-related factors on the level and mixture modeling;
growth of social skills, in a large sample of normally developing children in Norway (N = 2,076). On longitudinal; social skills
average, children’s social skills scores increased slightly, girls received higher scores than boys,
and individual order stability was fairly high over time. Growth mixture modeling (GMM) identified
three distinct trajectory classes, one with stable average scores over time (72%), and two (14%
each) with high initial and declining scores and low initial and increasing scores, respectively. The
school-related factors (e.g., student–teacher relationships) predicted social skills development
differently within the trajectory classes.

Social skills are life skills (UNICEF, 2012) and therefore school, adjust more smoothly to the student role, and have
crucial for children to acquire. Children who cooperate better grades than their less socially skilled peers (e.g.,
and share with others, who are helpful and empathic, and Hamre & Pianta, 2001; Zsolnai, 2002). Low social compe-
who are able to regulate emotions in adaptive ways, are tence has been associated with an array of unfavorable
generally better off on most social arenas in life, including at outcomes, such as externalizing behavior, delinquency,
school (e.g., Hamre & Pianta, 2001; Zsolnai, 2002). Social depression, social anxiety, academic failure, and unemploy-
skills are observable indicators of the larger social compe- ment (e.g., Kokko et al., 2006; Malecki & Elliot, 2002;
tence construct. Social competence is a multidimensional Masten et al., 2010; Obradović et al., 2010; Sørlie et al.,
construct, which refers to the ability to integrate cognition, 2008).
affect, motivation, and behavior in order to succeed with The acquisition of social skills is an essential develop-
social tasks and to achieve positive developmental out- mental process by which children from infancy learn to act
comes (Ogden, 2015). Most definitions converge that social and respond appropriately in social interactions and to
competence encompasses several related inter – and intra- form and maintain healthy relationships with others
personal skills that can be organized into overt sub- (Ogden, 2015). As children grow and gradually expand
dimensions, such as cooperation, assertion, self-control, their social environment, school becomes an important
empathy, and responsibility (Gresham & Elliott, 1990). arena where children both learn and exercise social skills.
Social skills are learned and affected by the characteristics At school, children’s social skills are challenged and shaped
of the context in which they develop. In this study, we in a variety of ways, both in dyads, small and large groups,
examined the degree to which school-related factors influ- and during interactions with peers and with teachers.
enced the initial level and growth of social skills in elemen- Despite evidenced relations to critical child outcomes,
tary school children, namely, student–student relationship, and in contrast to academic skills, social skills are seldom
teacher–student relationship, level of problem behavior in systematically promoted in schools (OECD, 2015). For
class and teachers’ collective efficacy. schools and teachers to be successful in supporting their
Social skills are important in their own right but have students, knowledge of how social skills develop and how
also been found to relate to other important domains of school may influence students’ social development is essen-
development, such as mental health (Humphrey & tial. In this study, we examined the development of social
Wigelsworth, 2012), coping (Bijstra & Jackson, 1998), and skills in a large sample of children from 4th through 7th
academic achievement (M. Welsh et al., 2001). Socially grade, and investigated whether a set of school-related
skilled children often have more positive attitudes toward factors affected children’s different growth trajectories.

CONTACT Mari-Anne Sørlie [email protected] The Norwegian Center for Child Behavioral Development (NUBU), P.O.BOX 7053 Majorstuen, Oslo
0306, Norway
© 2020 The Author(s). Published with license by Taylor & Francis Group, LLC.
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-
nd/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built
upon in any way.
S70 M.-A. SØRLIE ET AL.

Social skills development Nevertheless, our review indicates that results from
prior studies using well-validated and multidimensional
According to the bioecological model (Bronfenbrenner
measures of social skills (e.g., SSRS) report an increase
& Morris, 2007), proximal processes (i.e., interactions
during middle childhood (Berry & Connor, 2010;
between individuals in the microsystem such as family,
DiDonato, 2014), while results from studies using unidi-
school, and peers) are the primary engines of human
mensional measures (e.g., helpfulness) or different mea-
social development. Viewed from a transactional per-
surements at different waves, yield stable or decreasing
spective, there is a bidirectional interchange between
social skills scores (Côtė et al., 2002; Kokko et al., 2006;
children and their social contexts, in which both chil-
Nantel-Vivier et al., 2009; Obradović & Hipwell, 2010;
dren themselves and the environment develop and
Obradović et al., 2006; 2010; Sallquist et al., 2009). In the
change from the ongoing interactions. These transac-
current study, we used SSRS (Gresham & Elliott, 1990) to
tional processes may positively or negatively reinforce
measure children’s social skills across five time points
children’s social skills development, depending on the
from 4th grade through 7th grade.
contribution from the children themselves and the
environment (Sameroff & Mackenzie, 2003). In order
to foster children’s development of social skills, the Different developmental trajectories
proximal processes should be predominately positive,
When studying the development of children’s social
consistent, and supportive, and occur on a regular basis
skills, it is important to note that children vary with
over time (Davis, 2003).
regard to individual characteristics (e.g., tempera-
From theory, it can be expected that social skills ment, gender, cognitive abilities) and are differentially
gradually evolve through childhood and adolescence susceptible to environment characteristics (Belsky
(Bandura, 1986; Beauchamp & Anderson, 2010). In et al., 2007), including the school context. Some chil-
middle childhood (6 to 12 years), establishing and dren may be more sensitive and responsive to envir-
maintaining close friendships with peers are critical onmental experiences than others, and some may be
developmental tasks, as are learning and adapting ade- more easily influenced by peers than by teachers.
quately to the rules of school and society (Eccles, 1999). Moreover, environments may vary with regard to
Growth is dependent not only on the social learning how facilitative they are to such individual differences
opportunities afforded at home through modeling, (Iarocci et al., 2007). Children hold different experi-
reinforcement, and imitation (Bandura, 1997) but also ences from social interactions with parents, siblings,
on school-related factors such as relationships with and others, which will influence how they interact
peers and teachers (Ogden & Hagen, 2018). Hence, it with teachers and peers. The degree of parental sup-
is reasonable to expect that children’s social skills, in port, encouragement, and guidance may explain some
general, will be refined with practice, from social inter- of the variation in school-aged children’s social skills.
actions and reactions from the environment in various Children with well-developed social skills may more
situations. Given these assumptions, one may expect readily enter into positive relationships with teachers
that children with high scores on social skills measures and peers, which in turn reinforce their social skills.
at one time point will continue to score high on later Conversely, children who act inappropriately or awk-
measurement points (showing individual order stabi- wardly toward others may more often experience con-
lity), and that children on average would show frontation or rejection and thus miss out on positive
increases in social skills during the elementary school interpersonal experiences. From this, one might
years (showing a mean-level increase), rather than con- expect the initially more socially skilled children will
tinuity (Bornstein et al., 2017). advance their skills the most. On the other hand,
Despite the rich theoretical basis for advances in chil- children whose initial social skills are inadequately
dren’s social skills from toddlerhood to adolescence, the developed may have more room to grow and if they
empirical evidence of such growth is inconsistent (e.g., are provided with facilitating opportunities to interact
Carlo et al., 2007). A review of relevant literature revealed with peers and teachers, these children may, in fact,
nine longitudinal studies published since the turn of the exhibit greater growth.
century in which children’s social skills development Taken together, individual and contextual variability
across middle childhood was reported. As can be seen may through ongoing transactional processes contri-
from study characteristics and key results depicted in bute to differences in children’s social skills develop-
Table 1, results are diverging, and may stem partly from ment. It is reasonable to expect that groups of children
discrepancies in methods, which makes it difficult to draw may follow heterogeneous social skills trajectories and
conclusions from the literature. that these trajectories may be predicted by individual
Table 1. Study characteristics and key results from longitudinal studies (since 2000) of social skills development in children across middle childhood.
Study Instrument Subscales Raters Sample Psychometrics Longitudinal results
Berry and Connor (2010) Social Skills Rating Scale (SSRS; Total raw score of 30 items (e.g., sharing, Mothers N = 1364 normative US Cronbach’s alphas Mean level growth. No report
Gresham & Elliott, 1990). initiate friendship, controlling temper). sample from kindergarten range.87 to.91 across 6 on rank order.
through grade 6. time points. Four trajectories based on ratings
of internalizing and teacher-
child relationships.
DiDonato (2014) Interpersonal Skills scale from Mean score of five items (e.g., express Teachers N = 6610 normative US Cronbach’s alphas Mean level growth.
Social Rating Scale (SRS, adapted themselves positive, show empathy, help or sample from kindergarten range.88 to.89 across 5 No report on rank order.
from SSRS Gresham & Elliott, comfort peers). through grade 5. time points. Two trajectories, one higher level
1990). and one primarily stable,
moderate level with a slight
increase.
Côtė et al. (2002) SBQ (Social behavior Sum score of 10 items (e.g., helps hurt child; Teachers N = 1865 representative Cronbach’s alphas Mean level and rank order
questionnaire, Tremblay et al., praises others; shows sympathy). Canadian sample from range.89 to.91 across 7 stability within trajectory
1991), Helpfulness scale. kindergarten through time points. groups.
grade 6. Three trajectories, one high, one
medium, and one low.
Nantel-Vivier et al. (2009) Prosocial Behavior Scale (Caprara Mean score of six items (e.g., spends time with Teachers N = 472 normative Cronbach’s alphas Mean level stability or decrease
& Pastorelli, 1993). friends; helps others; shares things with Italian sample from 10 to range.85 to.91 across 5 within trajectory groups. Rank
friends). 14 years time points. order stability.
Four trajectories (low/stable,
moderate/declining, high/
declining, increasing).
Kokko et al. (2006) SBQ (Social behavior Sum score of 10 items (e.g., helps hurt child; Teachers N = 1025 low SES Cronbach’s alphas Mean level decrease, rank order
questionnaire, Tremblay et al., praises others; shows sympathy). Canadian boys from 6 to range.87 to.92 across 4 instability.
1991), Prosocial Behavior Scale. 12 years time points. Two trajectories, one moderate
and one low level with slight
decline.
Sallquist et al. (2009) Adapted version of Perceived Mean score of 4 items (e.g., this child is usually Teachers N = 199 US middle class Cronbach’s alphas Mean level decline and rank
Competence Scale for Children well behaved). sample from range.68 to.92 across 4 order stability.
(Eisenberg et al., 1995). kindergarten to early time points.
adolescence.
Obradović and Hipwell Perception of Peers and Self- Latent variable from 2 subscales (15 & 8 items) Self- N = 622 urban sample of Cronbach’s alphas Significant longitudinal stability
(2010) Inventory (POPS; Rudolph et al., (e.g., friends help you when you are in trouble; report US girls from 10 to range.67 to.78 across 4 between and across all annual
1995). kids like me because I am a good friend). 14 years time points. assessment waves.
Obradović et al. (2006) Latent measure of social Modified indicators with age (e.g., peer Multi N = 191 at risk US sample Cronbach’s alphas Mean level and rank order
competence; developmentally relationship, social skills, social competence). informant from kindergarten through range.87 to.92 across 4 stability across middle
appropriate multi method 16 years. time points (1st through childhood assessment waves.
indicators. 6th grade).
Obradović et al. (2010) Social competence measures from Subtly different indicators at each time point Multi N = 205 normative US All constructs show strong No report on mean level
broad domain (i.e., peer (i.e., specific indices vary with age). informant sample from 8 to 12 years. psychometric properties development.
acceptance, social skills, across 4 time points. Rank order stability across
friendships). assessment waves.
Out of space and comparison considerations, we focused on teacher reports when studies had multiple reporters for social skills. For the same reasons, when studies covered a longer time span, we focused on the
INTERNATIONAL JOURNAL OF SCHOOL & EDUCATIONAL PSYCHOLOGY

middle childhood years (6–12) when possible.

S71
S72 M.-A. SØRLIE ET AL.

and contextual factors in different ways. From the warrant further research, including additional explora-
bioecological perspective, it also derives that tion of mean level continuity and individual order
a longitudinal person-context approach is warranted consistency in social skills across middle childhood
when studying the development of children’s social (Carlo et al., 2007; Fraley & Roberts, 2005). Further
skills. research on the heterogeneity in children’s social skills
Empirical reports support that groups of children development is important from a preventive perspec-
may follow different trajectories in their social skills tive (e.g., to make possible early identification of chil-
development (cf., Table 1). For example, Côtė et al. dren at risk of social skills deficits, to guide the
(2002) observed from teacher-ratings of children’s development and implementation of effective interven-
helpfulness (SBQ), three trajectory groups for both tions to prevent social exclusion, to substantiate the
boys and girls (age 6 to 12): 1) a low trajectory group need for a social skills curriculum in school). In light
included children with the lowest helpfulness scores of the many negative impacts low social competence
and was stable across ages, 2) a moderate trajectory may have on children’s functioning, early identification
group included most children (54.4% of 930 boys, of students at risk of an unfortunate social skills devel-
58.2% of 937 girls) and had a curvilinear slope in opment is important. Likewise, knowledge on how tea-
which helpfulness first increased and then slightly chers best can support children on various social skills
decreased from age nine, and 3) a high trajectory trajectories would be of practical significance. By
group comprised of children with the highest helpful- reporting on the development of social skills in
ness scores which was stable across ages. More of the a large sample of typically developing children in
boys (43.8%) than of the girls (15.2%) were in the low Norway, the present longitudinal study aims to add to
trajectory group, while the reverse was true for the high the existing knowledge base.
trajectory group (1.8% of boys, 26.7% of girls). Nantel-
Vivier et al. (2009) identified four trajectory groups
from teacher-ratings of Italian children’s social skills
School-related factors that may influence the
(age 10 to 14) using the Prosocial Behavior Scale (low-
development of social skills
stable 8%, moderate-declining 48%, high-declining
37%, and increasing 7%). DiDonato (2014) found, Because children’s social functioning and development
based on teacher-ratings of social skills (SRS), two dis- of social skills are affected by the context in which they
tinct trajectories for both girls and boys (kindergarten develop, exploring the influence of factors related to the
through grade 5): a higher-level trajectory with school context is essential. Prior research indicates that
a marginally significant curvilinear shape and a stable it is the modifiable aspects of the school context (e.g.,
moderate-level trajectory. The high trajectory group student–teacher relationships, the policy and practice of
included most children (about 70%). the teachers as a group, extent of classroom disruption)
Results from these and other longitudinal studies rather than structural aspects (e.g., size, location, por-
indicate that the broad pattern of social skills develop- tion of males versus females) that affect students’ aca-
ment are similar for boys and girls. For example, in demic and behavioral outcomes (e.g., Hattie, 2009;
their longitudinal study of at-risk children, Obradović Sørlie & Torsheim, 2011; Welsh, 2003).
et al. (2006) found that social skills development School-related factors, such as the psychosocial
appeared gender invariant, in terms of both structure classroom environment may have differential effects
and stability, from early childhood to middle adoles- on children’s social skills trajectories. Classrooms char-
cence. There is, however, the extensive evidence of acterized by positive relationships between fellow stu-
gender differences in children’s level of social skills. dents likely create contexts more conducive to positive
Boys frequently receive significantly lower mean scores skills development. Spivak and Farran (2016) found
than girls, apparently regardless of the informant (tea- that children in preschool classrooms with more posi-
cher, parent, self, peers), age, and cultural context (e.g., tive and cooperative interactions between peers showed
DiDonato, 2014; Mpofu et al., 2004; Sørlie et al., 2008). greater gains in positive social behavior in 1st grade.
In summary, results from prior research on the While positive peer relationships in the classroom likely
development of social skills during middle childhood are generally beneficial, they may nevertheless be more
are mixed. Nevertheless, findings support theory in that important to certain subgroups of children than to
there are likely separate groups of children who follow others. For example, positive classmate relationships
different developmental trajectories. The number of may be of greater importance to children who show
trajectory groups that has been found to vary across a declining social skills trajectory than to those who
studies (most often 2–4). The diverging findings show an increasing or stable trajectory.
 INTERNATIONAL JOURNAL OF SCHOOL & EDUCATIONAL PSYCHOLOGY S73

Furthermore, the level of classroom problem beha- students and teachers. For example, students may rate
vior likely also affects children’s skills development their relationship with their teacher high on closeness,
differently. Kellam et al. (1998) found an interaction whereas the teacher may view the same relationship as
effect between disruptive classrooms and boys’ aggres- overly dependent on the part of the student. Second,
sion; the effects of disruptive classrooms were greater the predictive value of student–teacher relationship
for boys who were more aggressive, compared to both likely also depends on the rater. Koomen and Jellesma
the typical boy and aggressive boys who were placed in (2015) found no correlation between student-rated clo-
less disruptive classrooms. No similar classroom effect seness and teacher-rated prosocial behavior in their
was found for girls. It may be that children on a stable study of Dutch 4th to 6th graders. Finally, the student–
social skills trajectory are less susceptible to classrooms teacher relationship may have differential effects on
marked by frequent problem behavior or schools where students’ behavior depending on student characteris-
the teachers experience low collective efficacy than stu- tics. For example, Zee et al. (2013) found that tea-
dents on a less common trajectory. Thus, to the extent cher–student relationships were both closer and more
that children’s social skills development follow different conflictual among extroverted students. From theory
trajectories, it is likely that certain school-related pre- and prior research, it follows that the quality of rela-
dictors affect typical (majority growth curve) and aty- tionships between students and teachers is a highly
pical trajectories (growth curves with fewer students) relevant school-related factor with significant, but vary-
differently. ing influence on students' social functioning. In the
Research concerned with school-related factors that current study, we investigated whether student-rated
may affect the development of social skills has been student–teacher relationship quality predicted differen-
meager, especially during middle childhood (OECD, tial social skills trajectories.
2015). To be able to more effectively support all stu-
dent’s social development, teachers need explicit Student–student relationships
knowledge on how various school-related factors may From an early age, positive peer relationships and suc-
influence the student’s social skills. In this study, we cessful play interactions are associated with favorable
explored whether and how student-student and stu- social, behavioral and academic development, and
dent–teacher relationships, problem behavior in class adjustment (e.g., Wentzel et al., 2010). Conversely, con-
and teachers’ collective efficacy affected the students’ flictual peer interactions are associated with negative
social skills in 4th grade through 7th grade. behavioral and emotional outcomes, including difficul-
ties with school adjustment (e.g., Ladd et al., 1996).
Student–teacher relationships Despite the extensive literature on the importance and
Evidence suggests that healthy student–teacher rela- influence of young students’ social relationships on
tionships predict children’s later level of social skills their social functioning and academic achievements,
(e.g., Berry & Connor, 2010; Hamre & Pianta, 2001; surprisingly little is known about the predictive value
Pianta & Stuhlman, 2004). Children who experience of student–student relationships and whether they
close and conducive student–teacher relationships are affect the growth of social skills over time. In the
likely to be more socially adaptive than classmates with current study, we investigated whether student-rated
less positive relationships with their teachers (Griggs student–student relationship quality predicted differen-
et al., 2009; Pianta et al., 1995). Conversely, student– tial social skills trajectories.
teacher relationships marked by dependency and con-
flict are associated with negative outcomes, such as Problem behavior in class
poor academic performance, emotional insecurity, pro- Safe and constructive classroom environments with
blem behavior, and negative school attitudes (e.g., Split a lower prevalence of problem behavior make it easier
et al., 2018). to both teach and reinforce positive skills for teachers
Traditionally, studies investigating student–teacher and to exhibit and practice them for students (Ogden,
relationships have relied primarily on teacher reports, 2015). Classrooms with higher levels of disruption may,
although recent research has drawn attention to stu- on the other hand, form a more challenging context in
dents’ own assessments of their relationships with their which students can exercise a wide specter of social
teachers (Koomen & Jellesma, 2015). This raises some skills (negotiation, assertion, cooperation, self-control,
important issues. First, agreement between teacher- etc.). It is, however, not clear how a classroom level
reports and student-reports is generally low, which variable such as the amount of problem behavior affects
may indicate differences in perception or an emphasis the growth of an individual level variable such as social
on different aspects of the relationship between skills. In this study, we investigated the association
S74 M.-A. SØRLIE ET AL.

between the prevalence of classroom problem behavior Q2. Do children’s social skills follow the same overall
and children’s level and growth in social skills as this developmental trajectory across time points, or will
has not been empirically investigated before. separate trajectory classes be meaningfully identified?

Collective efficacy in school Q3. Do child gender and school-related factors (stu-
Collective efficacy refers to the teachers’ shared beliefs dent-student relationship, student-teacher relationship,
about their combined capability to organize and execute problem behavior in class, collective efficacy) predict
courses of actions required to produce student success intercept and slope?
(Goddard et al., 2004). Goddard et al. (2000) argue that
the teachers’ mutual beliefs of efficacy will shape the Q4. Do the school-related predictors have differential
normative culture of a school and subsequently have effects on children’s social skills in potentially different
modulating effects on the teachers’ behaviors, and that trajectory classes?
this, in turn, affects student performance. Previous studies
have demonstrated a positive association between tea-
chers’ collective efficacy and students’ academic achieve-
Method
ment (e.g., Goddard et al., 2004). Moreover, a strong
inverse and reciprocal relation between teachers’ collec- Data were drawn from a multi-cohort effectiveness
tive efficacy and student problem behavior has been study in which students (grades 4–7) and school staff
established (Sørlie & Torsheim, 2011). From social cog- in 65 elementary schools across Norway completed
nitive theory and prior research, one might expect that questionnaires at six measurement points (T1-T6)
teachers’ collective efficacy also influences students’ social across five successive school years. T1 assessments
skills. However, due to a lack of research, we do not know. were collected in the spring of 3rd grade (2007), T2
Thus, in the current study, we tested whether teacher assessments in the fall of 4th grade and the remaining
collective efficacy is a predictor of children’s social skills assessments were collected each spring in 4th, 5th, 6th
development. and 7th grade. To be able to follow the same group of
In sum, few studies have investigated the predictive students over a longer time period, the current analyses
effects of school-related characteristics and students’ concentrate on the 4th graders of the original sample
social skills (OECD, 2015). The present study contributes (n = 2,076 of 8,017, age 9 years) who were followed
to new knowledge by including the aforementioned through 7th grade (age 12 years). The schools were
school-related factors as potential predictors of children’s implementing various programs to prevent student
level and growth of social skills during the latter part of problem behavior or to improve the learning condi-
elementary school. Building on theory and prior studies, tions (Sørlie & Ogden, 2015) and were randomly
we hypothesized that student-teacher and student–stu- invited to participate as intervention or comparison
dent relationships, problem behavior in the classroom schools. Those who volunteered were selected accord-
and perceived collective efficacy in school predict stu- ing to predefined exclusion and inclusion criteria
dents’ social skills development over time, although we (detailed description of the design, see Sørlie &
also expect that they will have different effects for chil- Ogden, 2014). The aim of the current study was, how-
dren on various developmental trajectories. ever, not to examine intervention effects.
Head teachers rated the students’ individual social
skills at five time points (T2-T6, not at T1). Head teachers
Research questions
have the primary responsibility for the students and are
Based on repeated teacher assessments of social skills in typically the most knowledgeable about the students.
2,076 typically developing children in Norway and Some students had a different teacher rating their social
using a growth mixture model approach (GMM), we skills from 4th to 7th grade. Problem behavior in class and
asked the following research questions: collective efficacy were rated by school staff 6 months
prior (T1) to the first assessment of social skills. Student-
Q1. Is there mean-level and rank-order stability in rated student-student and student–teacher relationships
children’s social skills scores from grade 4 through were assessed parallel to the first assessment of social skills
grade 7; that is, is the average social skills score for (T2). The standards of the Regional Committee for
the children as a group stable across time, and do the Medical and Health Research Ethics (REK South-East)
children maintain their relative position within the and of the Norwegian Social Science Data Services were
group across time? followed throughout the conduct of the study.
 INTERNATIONAL JOURNAL OF SCHOOL & EDUCATIONAL PSYCHOLOGY S75

Participants and procedures (1 = never to 4 = almost all the time). The scale taps
three underlying sub-dimensions: assertion (e.g., “The
About 50% of the students were girls and 6.4% had
student initiates conversations with peers”), self-control
immigrant background (i.e., students and/or parents ori-
(e.g., “The student receives criticism well”) and coop-
ginated from another country and cultural context), of
eration (e.g., “The student attends to the teacher’s
which most were first or second-generation immigrants
instructions”). The sum score was used in the analyses
from Asian or African countries. Moreover, 3.9% had
(ɑ = .95 at T2 – T6).
been referred to the school educational services and
1.5% had been referred to child welfare or mental health
Student-teacher and student–student relationships
services during the baseline year (3rd grade). Also, 4.9%
Students rated how they perceived the social relationships
received ongoing special education. In Norway, special
in class on the 22-item Classroom Climate Scale (CCS;
education is a statutory right for students who for various
Sørlie & Nordahl, 1998). The CCS measures the psycho-
reasons have problems following the regular classroom
social learning conditions in class and consists of two
instruction. Reasons typically include learning difficulties,
significant sub-scales that were included in the analyses;
developmental disabilities, behavioral problems, or visual
a student–teacher relationship scale (STR; 10 items) and
or hearing impairments. Most teachers were females
a student–student relationship scale (SSR; 12 items).
(80%), experienced (69% had worked at least 11 years in
Items are assessed on a 4-point scale ranging from 1
schools, range 2 – 20+ years) and 78.5% were aged
(does not fit) to 4 (fits completely). Item examples are:
between 25 and 55 years. The remaining teachers were
“The students in this class are good friends” and “The
56 years of age or older. The mean school size was 297
teacher encourages me if I strive with a task.” The CCS
students (range 77–780). Data on students’ socio-
composite and sub-scales have shown satisfactory psycho-
economic status (SES) were not collected. However, the
metric properties in prior studies in terms of reliability
student sample was drawn from a diverse sample of
alphas and distributional qualities (e.g., Sørlie & Nordahl,
Norwegian schools located in areas with population SES
1998). In the present study, the reliability alphas of the
varying from relatively low to high.
STR and SSR were ɑ = .77 and .82, respectively.
Questionnaires were completed during ordinary
class sessions (1–2 h). To standardize the assessment
Problem behavior
procedures, written instructions were given. Informed
Incidences of moderate and serious problem behaviors
and written consent from parents was obtained before-
in the classroom context were measured by school staff
hand. All parents received an informative consent letter
on the 20-item scale Problem Behavior in the
(satchel-mail) in Norwegian or in four of the most
Classroom Last Week (Grey & Sime, 1989). School
frequently spoken foreign languages; English, Urdu,
staff reported on how many times they had observed
Somali, and Bosnia. The consent was signed and
negative behavior incidents in the classroom during
returned via the school. The consent letter included
a random week at T1. Item examples are; “Running in
key information on the research study, privacy, and
corridors” and “Physical attacks on students.” School
safe handling of information. The school staff agreed
staff included both head teachers, special education
coincident with completing the questionnaires.
teachers, teacher assistants, and after-school personnel.
Response choices ranged from 1 = not observed to
5 = observed several times per day. The scale has
Measures
shown satisfactory psychometric properties and validity
Social skills in prior Norwegian studies (e.g., Kjøbli & Sørlie, 2008;
The teacher version (elementary level) of the Social Lindberg & Ogden, 2001; Ogden, 1998). In the current
Skills Rating System (SSRS; Gresham & Elliott, 1990) study, reliability alpha was .88.
was used to assess the students’ social skills (the later
version, SSIS, the Social Skills Improvement System, Collective efficacy
was not available by the initiation of the study). This Teachers’ collective efficacy (TCE) was assessed by
well-validated measure has been used with diverse sam- school staff on a 12-item revised version (Goddard,
ples, and studies have reported it to be a reliable, gen- 2001) of the Collective Efficacy Scale (Goddard et al.,
der-invariant and valid measure, also in Norway (e.g., 2000). The TCE assesses teachers’ beliefs in their joint
Elliott et al., 1988; Klasussen & Rasmussen, 2013; effort and competence in promoting student success.
Ogden, 2003). Head teachers rated how often the stu- A recent validity analysis of the TCE supports its use
dents engaged in positive social behaviors described in (Hukkelberg & Sørlie, in review). Items were rated on
30 items on a 4-point (originally 3-point) scale a 5-point scale ranging from 1 (never) to 5 (very often)
S76 M.-A. SØRLIE ET AL.

with ɑ = .82 at T1. Item examples are; “Teachers here variance was fixed to zero as an initial exploratory test.
are confident they will be able to motivate their stu- Determination of the best-fitting model (i.e., the number of
dents” and “Teachers in this school are able to get classes) was based on assessments of BIC, SSA-BIC,
through to difficult students.” Measures in English entropy, parsimony, and theoretical considerations.
were independently translated and back-translated by Third, we tested a conditional LCGA model with the best-
two senior researchers at the Norwegian Center for fitting model from step two, adding the hypothesized cov-
Child Behavioral Development. ariates. We then addressed model convergence issues by
testing whether parameter estimates were replicated using
the OPTSEED option in Mplus, which helps to rule out the
Attrition and missing data analyses
possibility of ending up with local solutions. Fifth, we
Two thousand and seventy-six children participated at specified a conditional model in a growth mixture context
one or more assessment points. At T2 2,026 (98%) (GMM), entering the covariates while also allowing for
children participated, at T3 1,845 (89%), at T4 1,799 individual variation within latent trajectory classes to be
(87%), at T5 1,724 (83%), and at T6 1,675 (81%). One- freely estimated, represented by random effects. Finally, we
thousand four-hundred and sixty children (70%) parti- tested whether the covariates had different effects on the
cipated at all assessment points. The main reasons for intercept and slopes for the different classes, by comparing
attrition were change of school, absence on that model with a model in which the effects of the covari-
assessment day, and that one school withdrew from ates were fixed to be equal across classes.
the study prior to T2. All models were tested with the
full sample and full information maximum likelihood
Results
(FIML) was used to estimate information for partici-
pants with missing data. Table 2 presents the means, standard deviations, and
Compared to children who were retained at T6, correlations between study variables. As can be inferred,
children who were lost received significantly lower the rank-order stability (inter-individual stability) of
social skills scores at the first assessment (T2), were social skills across time for the child group as a whole,
less likely to have Norwegian as the first language, was relatively high at successive assessments, ranging
had teachers who scored lower on collective efficacy from r = .58 to r = .76. With reference to the question
and higher on classroom problem behavior. There were of rank-order stability (Q1), this indicates that children
no significant differences between children who were generally maintain their position within the child-group
retained at T6 and those who were not in terms of from one time-point to the next. As can be expected,
gender proportion, scores on the assessment of stu- individual (rank) order stability was lower across the
dent–student relations, student–teacher relations, or entire period, with r = .39 from T2 to T6. At every
school size. Analyses comparing children who partici- measurement point, teachers scored girls’ social skills
pated at all time-points to those who participated at significantly higher than they did boys.’ With reference
four or fewer time-points showed that children with to the question of mean-level stability (Q1), the results
partial data were less likely to have Norwegian as first showed a slight increase in scores over the five measure-
language (9.4% versus 5.5%). ment points (mean scores from 85.19 in 4th grade to
88.80 in 7th grade), suggesting that, in general, children
received somewhat higher social skills scores over time.
Analytic plan
To examine the longitudinal data for students’ social skills
Step 1: unconditional single-class linear and
from 4th through 7th grade, measured at five points, a series
quadratic growth curve models
of mixture models were tested, as recommended by Jung
and Wickrama (2008). Mixture models refer to a type of The unconditional linear growth model for social skills
analyses in which group membership of individuals is not fits the data reasonably well χ2 (6) = 11.46, p = .08,
known but is rather probabilistically inferred and repre- RMSEA = .02 (90% confidence interval, CI, = 0.00–
sented by categorical latent variables, known as trajectory 0.04) and CFI = .99. Both the latent intercept factor
classes (Muthén & Muthén, 2007). We built our models in (α = 84.50, t = 278.59, p < .01) and latent slope factor
the following steps: First, an unconditional single-class (β = 0.21, t = 7.84, p < .01) were significantly different
growth curve model was tested. At this step, we tested from 0, the latter indicating that the sample of students,
both linear and quadratic effects. Second, we specified on average, increased their social skills over time.
latent class growth analysis (LCGA) models comparing Furthermore, the variances of both the intercept (σi
one, two and more class solutions, where within-class 2 = 137, t = 16.46, p < .01) and slope (σs2 = .85,
 INTERNATIONAL JOURNAL OF SCHOOL & EDUCATIONAL PSYCHOLOGY S77

t = 11.62, p < .01) were significant. With reference to

SSRS-X = Teacher ratings of child social skills at waves 2, 3, 4, 5, and 6 (not measured at wave 1). Skewness and kurtosis values for all variables were within the −2 to 2 range (except the student–teacher relations
Student–teacher Relationship
the first part of the research question Q2 (“Do chil-
dren’s social skills follow the same overall developmen-
tal trajectory across time points”), our results seem to

(4.41)
34.32

-
indicate that there likely were subgroups of children
following different trajectories. The intercept and slope
factors correlated significantly r = −0.34, p < .01, indi-
cating that the lower the students’ social skills scores at
baseline, the greater their growth in social skills over
Student–Student Relationship

time. We then tested the same model, but with the

addition of a third random effect, a quadratic slope
factor. The addition of the quadratic slope factor pro-
(4.95)
40.32

.57
-
duced a less well-fitting model χ2 (2) = 6.46, p = .04,
RMSEA = .03 (90% confidence interval, CI, = 0.01–
0.06) and CFI = .99. A χ2 difference test, comparing the
linear and quadratic models, indicated that the linear
model was preferable; the difference between the mod-
Teacher collective

els was non-significant (p < .05, with 4 degrees of free-

dom), the linear model was simpler (parsimony), and
(2.85)
55.92
efficacy

.05
.06
-

achieved better χ2-value and fit indices.

Step 2: comparing unconditional latent class

Problem Behavior in Class

models
Next, we tested unconditional latent class models with
(5.61)
34.37

−.17
−.11
−.02

two and more classes to determine the optimal number

-

of classes. A latent class growth analysis model assumes

no within-class variances of individuals. Models were
compared based on log-likelihood values, BIC-, SSA-
BIC-, AIC-values, and entropy. In general, the higher
Table 2. Means, standard deviations, and bivariate correlations for key variables.

(16.62)
SSRS-6
88.80

.01
.03
.18
.20

the entropy value, the better. Lower BIC-values and

-

higher log-likelihood values are also indicative of better

(15.30)
SSRS-5

fitting models. None of these should be used alone as

87.13

−.03
.70

.09
.17
.15
-

an index of fit, rather the combination of all fit-indices

together with parsimony and theoretical considerations
(14.60)
SSRS-4
86.57

−.06

guided our model selection. Based on these indices, the

.65
.58

.11
.19
.20
-

model with three classes was considered the best-fitting

model. The Vuong-Lo-Mendell-Rubin likelihood-ratio
(14.89)
SSRS-3
85.51

.58
.49
.44
.02
.10
.19
.19
-

test (LMR-LRT) and the Bootstrapped likelihood-ratio

test (BLRT) for the k − 1 classes versus k classes (3
(14.08)
SSRS-2
85.19

classes) were both significant at the p < .001 level,

−.03
.76
.53
.43
.39

.07
.23
.22
-

suggesting that the 3-class solution provided better fit

to the data than the 2-class solution. Results thus
Gender

−.02
% girls
50.2

.28
.29
.24
.21
.20
.03

.12
.18

seemed to affirm the second part of the research ques-

-

tion Q2 (“will separate trajectory classes be meaning-

fully identified?”).
Student-Student Relationship
Student-Teacher Relationship

assessment which was 3).

Teacher Collective Efficacy
Problem Behavior in Class

Step 3: conditional latent class growth model

Percent/Mean

We then tested a 3-class (from step 2) conditional

latent class model, entering the covariates: student–
(Std.dev)
Gender
SSRS-2
SSRS-3
SSRS-4
SSRS-5
SSRS-6

teacher relationship, student–student relationship, pro-

blem behavior in class, and teacher collective efficacy.
S78 M.-A. SØRLIE ET AL.

Child gender was entered as a predictor of class mem- predictors fixed to be equal, but with both intercept
bership. The model tested in this step was still in the and slope freed, was tested next. Although the entropy
latent class growth analysis framework, which fixes the value for this model was high, the remaining fit-indices
within-class variances to zero. This model produced were quite low. Next, we tested the same GMM 3-class
a sample division of 18% in class 1, 55% in class 2, model, but allowing the predictors to have different
and 27% in class 3. All trajectory classes showed rela- effects on the classes, and by successively freeing para-
tively flat curves for social skills over time but were meters of the models (GMMc – GMMg). As can be seen
different in their respective levels. All covariates were from Table 3, two models obtained better fit indices
significantly predictive of the intercept, in the expected (e.g., BIC, entropy) than the remaining models, that are
direction, with the exception of problem behavior in GMMa and GMMf. These two models were quite simi-
class, which was non-significant. Child gender was the lar with regards to their respective fit indices and they
only significant predictor of the slope factor, suggesting are nested; one has the constraints of holding the cov-
that boys were more likely to show increases in social ariates’ effects equal within classes, whereas the others
skills over time, compared to girls. Results from the let the effects of the covariates be freely estimated
analyses in this third step answered research question within classes. We performed a likelihood-ratio differ-
Q3 (“Do child gender and school-related factors predict ence test using their respective log-likelihood (LogL,
growth factors?”). When inspecting the graphics for the H0) values. The more restrictive model (GMMa) had
respective classes, however, it seemed reasonable to a LogL value of −29,843.64 with 28 parameters, whereas
assume that at least two of the classes (classes 1 the less restrictive model (GMMf) had a LogL of
and 3) needed their own class-specific variances. −29,791.45 with 52 parameters. This Log-likelihood
Nevertheless, before we tested such models we wanted difference was multiplied with two, representing the
to ensure that our results were not the product of local value to which the chi-squared critical value is com-
solutions. pared (difference of 24 degrees of freedom). The Log-
likelihood difference value exceeded the critical value at
α = .001, suggesting that the model with less con-
Step 4: addressing model convergence issues
straints, GMMf, has the better fit and thus we proceed
Prior to comparing LCGA and GMM models (GMM with interpreting the results from this model in the
models were tested in a step-wise fashion by progres- following. Figure 1 depicts the 3-class solution from
sively letting the variances of the growth factors to be GMMf, for which the intercept was freed, the slope
freely estimated), we addressed convergence issues. variance fixed at zero, and the effects of predictors
Although all the models tested thus far converged suc- were freed.
cessfully, we checked for the possibility of local solu-
tions. We re-ran the conditional latent class model
Results from GMMf
described above using the OPTSEED option in Mplus.
We used the two best loglikelihood values from the A summary of the effects of the predictors on each of
output and re-ran the model with the respective seed the trajectories’ growth factors is presented in Table 4.
values. If estimates are replicated in these two runs, the Furthermore, Table 5 shows the class percentage and
chances for ending up with local solutions are reduced. the estimated correlation between intercept and slope
Results showed that class proportions remained the and explained variance of social skills within each class.
same and all estimates were replicated, suggesting that The following presentation of results addresses research
we did not end up with local solutions. question Q4 (“Do the school-related predictors [and
gender] have differential effects on children’s social
skills in […] different trajectory classes?”).
Step 5: conditional growth mixture model with
The High-Declining (class 1) comprised about 14%
three classes
of the sample and was characterized by high initial
Finally, we examined a 3-class conditional growth mix- social skills scores (estimated mean = 97.92), followed
ture model entering the predictors in the overall model by a steady decline over time (slope = −1.52). The
(as was the case with the above conditional LCGM), estimated correlation between intercept and slope was
allowing for within-class variances. We first specified −.37, suggesting that the higher the intercept score, the
a model in which the effects of all the predictors on the less of an increase in social skills scores over time. The
growth factors were fixed to be equal, with intercept model explained from 50% to 75% of the variances in
freed and slope variance fixed at zero (GMMa). the social skills measure at the various time-points and
A second model (GMMb), also with the effects of 39% of the intercept variance.
 INTERNATIONAL JOURNAL OF SCHOOL & EDUCATIONAL PSYCHOLOGY S79

For the students in this class, initial social skills scores

LCGA = latent class growth analysis; GMM = general mixture modeling; BIC = Bayesian information criterion; SSA-BIC = sample size-adjusted Bayesian information criterion; AIC = Akaike information criterion;
Entropy
.75
.74

.95

.72

.73

.76

.74

.79
(intercept) were positively associated with student–tea-
cher relationship (.46, p < .001) and with being a girl
(.33, p < .001). Conversely, the greater the scores on
student–student relationship, the lower the intercept of

113,254.93
60,536.47
59,743.29

59,711.43

59,714.69

60,505.01

59,686.89

59,546.43
AIC

social skills (−.31, p < .01), and the greater the scores on
collective efficacy, the lower the initial scores on social
skills (−.34, p < .001). The covariate problem behavior in
class was unrelated to the intercept in this class. The only
113,387.83
60,593.43
59,807.08

59,820.79

59,828.61

60,616.65

59,805.37

59,674.02
SSA-BIC

covariate that significantly predicted the slope factor in

class 1 was student–student relationship. Greater stu-
dent–student relationship scores were associated with
less decline in social skills over time (1.03, p < .001).
113,559.39

The Moderate-Stable (class 2) comprised about 72% of

60,672.85
59,896.03

59,973.28

59,987.45

60,772.32

59,970.57

59,851.93
BIC

the sample and was characterized by average initial scores

on social skills (estimated mean = 84.18), followed by
a rather flat slope over time (slope = .18). The estimated
correlation coefficient between intercept and slope was
Number of parameters

−.18, suggesting that the greater the intercept score, the

less growth over time. The model explained from 61% to
25
28

54

48

50

49

52

56

78% of the variances in social skills at the various time-

points and 20% of the variance in the intercept.
All covariates predicted initial scores (intercept) in
this class; the greater the scores on student–student
relationship (.14, p < .01), student–teacher relationship
Log-likelihood (H0)

(.14, p < .001), collective efficacy (.12, p < .01), problem

−56,573.464
−30,243.23
−29,843.64

−29,807.72

−29,807.34

−30,203.50

−29,791.45

−29,717.21

behavior in class (.10, p < .05) and being a girl (.30,

p < .001), the greater the intercept scores of social skills.
The greater the scores on collective efficacy (.49,
p < .05) and the lower the scores on problem behavior
in the class (−.49, p < .05), the greater increases were
observed in social skills over time (slope). Greater scores
equal across classes, slope var. fixed at zero, predictors freed within each class)

on student–teacher relationship, however, were asso-

ciated with less increases in social skills (−.43, p < .05).
Gender also predicted slope in this class, being a boy was
related to greater increases in social skills (−.49, p < .01).
and slope held equal across classes, effects of predictors freed)

The covariate student–student relationship was unrelated

Var. = variance. * the best log-likelihood value was not replicated
and slope var. fixed at zero, pre-dictors freed within class)
freed, slope var. at zero, effects of predictors held equal)
Table 3. Model comparison of LCGA and GMM models.

freed, slope var. fixed at zero, effects of predictors freed)

to the slope factor in this class.

The Low-Increasing (class 3) comprised about 14%
and slope freed, effects of predictors held equal).

of the sample and was characterized by lower initial

and slope freed, effects of predictors freed)*

scores on social skills (estimated mean = 77.79), fol-

lowed by an increase over time (slope = 1.90). The
estimated correlation coefficient between intercept and
slope was .06, suggesting that the intercept score was
not associated with growth over time. The model
explained from 50% to 66% of the variances in social
skills measure at the various time-points and 37% of
the intercept variance.
For the students in this class, higher intercepts were
predicted by greater scores on student–student rela-
tionship (.27, p < .01) and with being a girl (.41,
(Intercept
(intercept

(intercept

(intercept

(intercept

(intercept

(intercept
GMMb

GMMd

GMMg
Model

GMMe
GMMa

GMMc

GMMf
LCGA

p < .001). The remaining covariates were unrelated to

intercept scores in this class.
S80 M.-A. SØRLIE ET AL.

110.00

100.00

90.00
SSRS
80.00
Score
70.00

60.00

50.00
T2 T3 T4 T5 T6
Time

Class 1 (High-declinig) Class 2 (Moderate-stable) Class 3 (Low-Increasing)

Figure 1. Estimated means across time for the three trajectory classes (GMMf).

Table 4. Parameter estimates of predictors on the different trajectories’ growth factors.

Trajectory class intercept Trajectory class slope
High-declining Stable-medium Low-increasing High-declining Stable-Medium Low-Increasing
Predictor (class 1) (class 2) (class 3) (class 1) (class 2) (class 3)
Student-Teacher .46** .14** n.s. n.s. −.43** −.57**
Relationship
Student-Student Relationship −.31** .14** .27** 1.03** n.s. .49*
Problem Behavior in Class n.s. .10* n.s. n.s. −.49** −.74**
Collective Efficacy −.34** .12** n.s. n.s. .49* n.s.
Gender .33** .30** .41** n.s. −.49* −.21*
*Significant at the .05 level. **Significant at the .01 level. Gender is coded boys = 1, girls = 2. Parameter estimates are standardized.

Table 5. Percent of sample within each group, estimated correlation between intercept and slope, and explained variance.
Trajectory class
High-declining (class 1) Stable-medium (class 2) Low-increasing (class 3)
Percent of sample 14% 72% 14%
Intercept and slope correlation −.37 −.18 .06
Explained variance in social skills 50%–75% 61%–78% 50%–66%

Moreover, the greater the scores on student–student Discussion

relationship, the greater the increases in social skills over
time (.49, p < .01). Likewise, less problem behavior in The current study examined the development of tea-
class predicted increases in social skills (−.74, p < .01) as cher-rated social skills in a large sample of elementary
did being a boy (−.21, p < .05), while greater scores on school children in Norway and the potential influences
student–teacher relationships were associated with less of child gender and school-related factors on the level
increase (−.57, p < .01). Teacher collective efficacy was and change in social skills. There was a slight mean-
unrelated to growth in social skills in this class. level growth in social skill scores from 4th through 7th
Taken together, the results from the analyses in steps 1 grade, and in general, children maintained their rank
to 5 indicated that children’s social skills do not follow the within the group from one time-point to the next.
same overall developmental trajectory across time (Q2). Three trajectory groups with distinct developmental
Rather, separate trajectory classes were meaningfully iden- pathways were identified, one with high initial and
tified, in that subgroups of children seemed to follow declining scores, one with low initial and increasing
different social skills trajectories over time (Q2). scores, and one with average initial and fairly stable
Moreover, the results indicated that both child gender social skill scores. The school-related factors tested in
and the school-related predictors affected the growth fac- this study predicted children’s level and development
tors (intercept and/or slope, Q3) and these effects seemed of social skills differently within the respective trajec-
to vary within trajectory classes (Q4). tory groups.
 INTERNATIONAL JOURNAL OF SCHOOL & EDUCATIONAL PSYCHOLOGY S81

General growth during middle childhood both Côtė et al. (2002) and Nantel-Vivier et al. (2009)
found a low, medium, and a high score group. In other
The observed general growth in social skills across
studies, a two-trajectory solution has fitted the data best
middle childhood is consistent with most theoretical
(DiDonato, 2014; Kokko et al., 2006).
models positing that social skills gradually increase
through childhood and adolescence (e.g., Bandura,
1986; Beauchamp & Anderson, 2010). The increase Trajectories and school-related predictors
also corresponds with findings from two earlier studies
Moderate-stable group
in the USA (Berry & Connor, 2010; DiDonato, 2014).
The vast majority of the children followed the
The empirical basis of growth in social skills from
Moderate-Stable trajectory of social skills development.
toddlerhood and onward derived from the relative few
As expected, for most children, higher social skills
studies that do exist, is, however inconsistent. Indeed,
scores at the beginning of grade 4 were predicted by
the lion’s share of prior studies have reported either
positive relationships with peers and teachers, and by
a slight decrease or general stability from the beginning
being a student in a school where the staff experienced
of middle childhood to early adulthood (Côtė et al.,
greater collective efficacy.
2002; Kokko et al., 2006; Nantel-Vivier et al., 2009;
Perhaps more surprising was that higher rates of
Obradović et al., 2010; Obradović & Hipwell, 2010;
problem behavior in the classroom in 3rd grade also
Obradović et al., 2006; Sallquist et al., 2009).
predicted these children’s initial social skills scores in
Mixed findings may partly stem from differences in
4th grade (intercept). Prior research has shown that
measures. Notably, few prior studies have used social
social skills and problem behavior are not mere oppo-
skills measures that take into account the multidimen-
sites (e.g., Sørlie et al., 2008). Said differently, socially
sionality of the construct or have used well-established
skilled students may also display some unruly behavior,
and validated measures. Mixed findings may also stem
and children who are less socially skilled can be quiet
from sample differences; small sample sizes, US versus
and reserved and thereby contribute to a more reticent
European samples (e.g., Sallquist et al., 2009), single-sex
classroom environment. Thus, the question becomes,
samples (e.g., Obradović & Hipwell, 2010) and samples
why does an aggregated measure of problem behavior
with different socio-economic backgrounds (e.g.,
in the class predict greater initial social skills scores for
Kokko et al., 2006). Moreover, discrepancies may
the majority of children in this sample? One explana-
reflect variations in the raters’ knowledge, observations,
tion may be that a few students only are responsible for
or relations to the child.
disruptive behaviors and compared to them, most other
students appear (and indeed probably are) more
socially competent. Second, in classrooms with frequent
Social skills trajectories
problem behaviors, most children may learn more
Our longitudinal data suggest that there is heterogeneity quickly to navigate socially and find ways to relate to
among children in their social skills development and that the student role and classmates in adaptive ways. It may
children can be meaningfully grouped into distinct tra- also be that in classrooms characterized by frequent
jectory classes. A model comparison process revealed that incidents of disturbing behaviors, teachers soon realize
a 3-class solution fits the data best. A closer examination a need for clear behavior expectations and proactive
of the finding that children’s social skills generally seem to disciplinary practices in order to reduce disruption or
increase slightly over time revealed that, in fact, most to promote the students’ social skills.
children (72%) follow a Moderate-Stable trajectory In line with expectations, higher teacher collective
(class 2). Perhaps more surprising and important for efficacy and low prevalence of classroom problem beha-
teachers and interventionists, we also found that some vior predicted greater growth in social skills over time
children (14%) follow a High-Declining trajectory (slope) in this trajectory class. Thus, it seems that
(class 1), while an equally small group (14%) follows growth in social skills in the later grades of elementary
a Low-Increasing trajectory (class 3). The High- school is more likely to occur in more orderly class-
Declining group may be considered at high risk of social rooms and in schools where the teachers report
exclusion and other negative life outcomes. a greater common faith in their capacity to support
Other studies have also found that there is hetero- students. Contrary to expectation, growth in social
geneity in children’s social skills development during skills in this group also was negatively predicted by
the elementary school-years (e.g., Côtė et al., 2002; student-rated student–teacher relationship. This may
DiDonato, 2014; Kokko et al., 2006; Nantel-Vivier reflect a tendency of teachers and students to place
et al., 2009). For example, as in the present study, emphasis on different aspects of their relationship or
S82 M.-A. SØRLIE ET AL.

that students who early on feel close to their teachers teachers’ socials skills ratings over time, this subgroup
have more trouble shifting their relationship-focus to may alternatively be spoken of as “late bloomers” or
peers and thus appear less socially skilled. The correla- “social survivors.” The labels reflect that although they
tion between the slope and intercept was non- by grade 4 are perceived by their teachers as the socially
significant in this group. less skilled, their social skills improve as they gradually
experience social success, and the more socially skilled
High-declining group they become, the more positive reinforcement they get.
A negative correlation between the intercept and slope In other words, gaining social competence is rewarding
indicated that students in High-Declining group with (Sørlie et al., 2008), and at the end of elementary
the highest social skills scores by grade 4 showed the school, the teachers see these students as highly socially
greatest decline over time. Positive student–teacher skilled children who enjoy making friends and being
relationships (student-rated) predicted greater social socially active in class.
skills scores in grade 4 (intercept) in this trajectory As for the Moderate-Stable group, more positively
group, while greater student–student relationship pre- rated relationships with classmates predicted higher
dicted lower initial scores. These results suggest that initial intercept scores in the Low-Increasing group.
a child in this trajectory group who rate the relation- Moreover, positive peer relationships paired with
ship with the teacher as positive feels supported by and a classroom climate with lower rates of problem beha-
dependent on her teacher, and that she may seek the vior predicted greater growth (slope) in these students’
company of and support from teachers at the expense social skills over time. On the other hand, a positive
of cultivating friendships with classmates and peers. student–teacher relationship score was associated with
Although many teachers initially may perceive a child less increase in social skills scores over time for this
that frequently seeks adult acknowledgment and atten- trajectory group as well.
tion as socially competent, the negative trend in the These results suggest that for the children in both of
social skills ratings over time suggests that they gradu- these trajectory classes – and in accordance with our
ally change their minds. As these children move on to hypotheses on what went on in the High-Declining
5th, 6th and 7th grade, the teachers may come to see group, a relationship with the teachers which the chil-
them as clingy or socially helpless rather than as dren perceived as highly positive, seemed less condu-
socially competent. Another plausible explanation is cive to growth in social skills over the latter part of the
that teachers gradually become aware of these chil- elementary school-years. This apparently contra-
dren’s lack of peer involvement and rate their social intuitive finding corresponds, however, with findings
skills accordingly. from some prior studies using a more fine-meshed
Lower teacher collective efficacy scores also pre- measure of student–teacher relationships, indicating
dicted a higher level of initial social skills in this trajec- that when children are too dependent on their teachers,
tory class. Perhaps in schools where there is little it is not beneficial to building their interpersonal skills
consensus about the teachers’ joint capability to pro- (e.g., Berry & Connor, 2010; Hamre & Pianta, 2001;
mote students’ social functioning, teachers are less in Pianta & Stuhlman, 2004).
a position to identify “warning sign” of unfortunate Taken together, the school-related predictors tested
social skills development, such as a student’s overly in this study affected children’s social skills in the
reliance on teachers. three trajectory classes differently. The role of rater
The only factor that predicted the slope in the High- (student versus teacher) seems particularly important
Declining group was student–student relationships. to consider in this regard. Our analyses revealed that
This finding suggests that the general downward slope student–student and student–teacher relationships as
in this trajectory class is somewhat less steep for chil- rated by the students themselves stood out as more
dren who score more positively on student–student influential than the level of problem behavior in class
relationships in 4th grade. Thus, it seems that if there and teacher’s perceived collective efficacy. While
is anything that can “save” these “teacher’s pets” from extensive empirical work and a widespread popular
becoming gradually less socially competent, it is getting opinion place a high value on the student–teacher
along with their classmates early on. relationship, underlining its importance for students’
learning and well-being (e.g., Hattie, 2009), our
Low-increasing group results suggest that one needs to look more carefully
There was a non-significant association between the at both the positive and potentially negative impacts
slope and intercept in the Low-Increasing group. In of student–teacher relationships on children’s’ social
accordance with the evidently positive trend in the skills development. From the study results, it seems
 INTERNATIONAL JOURNAL OF SCHOOL & EDUCATIONAL PSYCHOLOGY S83

that for most students, a proper and competent class- relationship, problem behavior in class, and teacher
room management accompanied by rich opportu- collective efficacy influence the development of social
nities for positive relationships with peers is more skills in different groups of children. However, several
favorable to their social skills development than one- other important influences are likely at play in shaping
to-one interactions with teachers. children’s social skills development (e.g., parenting
practices, social training, classroom leadership).
Future research could expand on the current study by
Gender differences
including additional predictors.
In line with evidence from numerous studies (e.g., Other strengths of the study are the inclusion of
DiDonato, 2014; Eiden et al., 2009; Sørlie et al., 2008), a large sample of typically developing elementary
the current results showed that girls on average scored school children and the use of a well-validated and
higher on social skills than did boys at all measurement comprehensive measure of social skills across five
points. The gender differences were, however, greater time points. Moreover, to reduce bias stemming from
for the initially less socially skilled children (Low- mono-informant data, we used cross-rater assessment
Increasing) than for the initially more skilled children and hereby strengthening the validity of the study
(High-Decreasing). Thus, socially skilled boys and girls results (Shadish et al., 2002). On the other hand, the
appear more alike, whereas socially unskilled boys and children’s social skills were rated in one context and by
girls seem more different, with boys scoring consider- teachers only, which may be noted as a limitation.
ably lower than girls. Although teachers have been found to be reliable raters
Such gender differences have been tentatively of children’s social skills (Ogden, 2003; Renk & Phares,
explained by teachers’ biased gender expectations regard- 2004; Sørlie et al., 2008), adding self-reports, direct
ing socially competent behavior. That is, teachers may observations or parent-reports would have increased
anticipate boys to show less adaptive behavior and girls to the validity of the findings. Moreover, there is
behave more compliant (Parks & Kennedy, 2007). a newer version of the social skills scale, the SSIS
However, this explanation seems inadequate since boys (Social Skills Improvement System, Gresham & Elliott,
generally receive significantly lower scores than girls 2008) than the one used in the current study. However,
regardless of informants or context (e.g., DiDonato, SSIS was not available at the time the study was
2014; Eiden et al., 2009; Ogden, 2003; Sørlie et al., initiated, and change of measure across assessment
2008). It may be the case, as others have suggested, that points was not seen as adequate. We preferred the
girls’ higher social skills ratings reflect earlier maturity, SSRS version also to enable comparisons of the results
greater self-regulation, and social adaptability compared with those of earlier studies.
to boys (e.g., Westiling et al., 2012). We did not assess the students’ level of risk, which
Few prior studies have examined whether the gen- may be an important precursor or covariate in models
erally observed difference in the level of social skills such as ours. Furthermore, the assessment of social
favoring girls decrease, increase, or is stable across skills started in 4th grade and by this point in time
years. Côtė et al. (2002) found no gender differences teachers’ perceptions of students’ social skills may be
in the development of helpfulness skills during elemen- fairly well established, in the sense that teachers may
tary school. A multiple group analysis by Westiling view and therefore rate children based on earlier opi-
et al. (2012) showed that girls had a higher mean nions about and experiences with their students.
intercept and slope of social skills than boys across Besides, some unexpected findings emerged. For exam-
grades 6 to 8. Interestingly, in two of the latent trajec- ple, in both the Moderate-Stable and Low-Increasing
tory classes in the current study (Moderate-Stable and trajectory, higher student–teacher relationship pre-
Low-Increasing) being a boy was associated with dicted less growth in social skills over time (although
a steeper upward slope, perhaps indicating that boys student–teacher relationship did positively predict the
may “catch-up” over time. intercepts).
From the results, strengthening student–student
relationships seems especially important in order to
Strengths, limitations, and future directions
curb a decline or encourage an increase in children’s
Few studies have explored predictive relationships social skills over time. In particular, teachers ought to
between school-related characteristics and the develop- be attentive to students who are too close or too depen-
ment of students’ social skills (OECD, 2015). The pre- dent on them, although many of these students behave
sent study adds to the knowledge base by exploring well and rarely misbehave. Excessive reliance on adults
how student–student relationship, student–teacher may ʽback-fire’ on the children and render them
S84 M.-A. SØRLIE ET AL.

helpless or inexperienced in peer interactions. For some large-scale longitudinal studies are necessary before
students, problem behavior in the classroom also seems conclusions can be drawn.
to be a risk factor for their development of social skills.
Perhaps classroom environments characterized by
noise and disorder rob certain children of opportunities Acknowledgments
to exercise pro-social skills. A well-led and orderly This research was supported by the Norwegian Center for
classroom is not only conducive to academic learning Child Behavioral Development (NUBU) and in part by
(e.g., Korpershoek et al., 2016) but also models appro- a grant from the Norwegian Directorate for Education and
priate behavior and communicates rules for acceptable Training. We are thankful for the data support from Asgeir
and positive behavior. Clearly, it would be of great Røyrhus Olseth.
importance for future studies to see whether the
observed associations between school-factors and social
Disclosure statement
skills trajectories can be replicated, preferably in long-
itudinal studies using reliable measures of social skills. No potential conflict of interest was reported by the authors.
Moreover, we did not test the interaction effects of the
predictors and so the potential for one predictor (e.g.,
Notes on contributors
student–teacher relationship) to affect social skills devel-
opment depending on levels in another predictor (e.g., Mari-Anne Sørlie is a researcher at the Norwegian Center for
problem behavior in class) remains untested. It would be Child Behavioral Development. Interests: effectiveness of
interesting in future studies to examine such interactions, school interventions, behavior problems, social competence,
implementation fidelity. Author of several articles and book
for example, problem behavior in class may exert more
chapters.
influence on children’s social skills development in cases
of lower relationship quality with peers. Finally, because Kristine Amlund Hagen is research director at the Norwegian
this study was conducted in schools delivering some kind Center for Child Behavioral Development. Interests: develop-
mental psychology, psychosocial difficulties, social compe-
of measures to increase a positive learning environment, tence, and intervention research. She has published several
caution must be exercised when interpreting the social research articles.
skills means across time points.It can, however, be noted
Kristin Berg Nordahl is a researcher at the Norwegian Center
that the social skills mean score at baseline corresponded
for Child Behavioral Development. Interests: observation as
well with the mean score in a sample of same-aged research method, early child-parent interactions, social com-
students from a representative sample of Norwegian petence and behavior problems in childhood.
schools (Ogden, 1995), indicating a normal sample and
that the study results are valid beyond the current sam-
ple, at least within the Norwegian context. Moreover, we References
re-ran the final model adding school program type as
Bandura, A. (1986). Social foundations of thought and action:
a predictor. It did not change the curves of the trajec- A social cognitive theory. Prentice Hall.
tories, the proportions of students in each trajectory, nor Bandura, A. (1997). Self-efficacy: The exercise of control.
the direction of effects of each of the other predictors. W. H. Freeman/Times Books/Henry Holt & Co.
This additional model testing supports the robustness of Beauchamp, M. H., & Anderson, V. (2010). Social: An inte-
our results. grative framework for the development of social skills.
Psychological Bulletin, 136(1), 39–64. https://doi.org/10.
1037/a0017768
Belsky, J., Bakermans-Kranenburg, M. J., & van
IJzendorn, M. H. (2007). For better and worse:
Conclusions Differential susceptibility to environmental influences.
The current study adds to the empirical foundation Current Directions in Psychological Science, 16(6),
200–304. https://doi.org/10.1111/j.1467-8721.2007.00525.x
showing that children’s development of social skills
Berry, D., & Connor, E. (2010). Behavioral risk, teacher–child
during middle childhood seems to follow relationships, and social skill development across middle
a differentiated and complex pattern. The most note- childhood: A child-by-environment analysis of change.
worthy findings indicate that both the level (intercept), Journal of Applied Developmental Psychology, 31(1), 1–14.
shape, and direction of children’s social skills trajec- https://doi.org/10.1016/j.appdev.2009.05.001
tories (slope) are partly dependent on qualitative and Bijstra, J., & Jackson, S. (1998). Social skills training with
early adolescents: Effects on social skills, well-being,
modifiable aspects of the school context. Particularly, self-esteem and coping. European Journal of Psychology of
the students’ relationships with their peers and teachers Education, 13(4), 569–583. https://doi.org/10.1007/
seem influential – for “better or worse.” However, more BF03173106
 INTERNATIONAL JOURNAL OF SCHOOL & EDUCATIONAL PSYCHOLOGY S85

Bornstein, M. H., Putnick, D. L., & Esposito, G. (2017). student achievement. American Educational Research
Continuity and stability in development. Child Journal, 37(2), 479–507. https://doi.org/10.3102/
Development Perspectives, 11(2), 113–119. https://doi.org/ 00028312037002479
10.1111/cdep.12221 Goddard, R. D., Hoy, W. K., & Woolfolk Hoy, A. (2004).
Bronfenbrenner, U., & Morris, P. A. (2007). The bioecological Collective efficacy beliefs: Theoretical developments, empiri-
model of human development. In R. M. Lerner (Ed.), cal evidence, and future directions. Educational Researcher,
Handbook of child psychology (Vol. 1, pp. 793–828). John 33(3), 3–13. https://doi.org/10.3102/0013189X033003003
Wiley & Sons, Inc. Gresham, F. M., & Elliott, S. N. (1990). Social skills rating
Caprara, G. V., & Pastorelli, C. (1993). Early emotional system. Manual. American Guidance Service.
instability, prosocial behaviour, and aggression: Some Gresham, F. M., & Elliott, S. N. (2008). Social skills improve-
methodological aspects. European Journal of Personality, ment system: Rating scales. Pearson Assessments.
7(1), 19–36. https://doi.org/10.1002/per.2410070103 Grey, J., & Sime, N. (1989). Findings from the national survey
Carlo, G., Crockett, L. J., Randall, B. A., & Roesch, S. C. of teachers in England and Wales. In Elton (Ed.), Discipline
(2007). A latent growth curve analysis of prosocial beha- in schools. Report of the committee of enquiry chaired by Lord
vior among rural adolescents. Journal of Research on Elton Department of education and science and the Welsh
Adolescence, 17(2), 301–324. https://doi.org/10.1111/j. office. Her Majesty’s stationary office (pp. 220–292).
1532-7795.2007.00524.x Retrieved from http://www.educationengland.org.uk/docu
Côtė, S., Tremblay, R. E., Nagin, D., Zoccolillo, M., & ments/elton/elton1989.html
Vitaro, F. (2002). The development of impulsivity, fearful- Griggs, M. S., Gagnon, S. G., Hulesman, T. J., Kidder-Ashley,
ness, and helpfulness during childhood: Patterns of con- P., & Ballard, M. (2009). Student–teacher relationships
sistency and change in the trajectories of boys and girls. matter: Moderating influences between temperament and
Journal of Child Psychology and Psychiatry, 43(5), 609–618. preschool social competence. Psychology in the Schools, 46
https://doi.org/10.1111/1469-7610.00050 (6), 553–567. https://doi.org/10.1002/pits.20397
Davis, H. A. (2003). Conceptualizing the role and influence of Hamre, B. K., & Pianta, R. C. (2001). Early teacher-child
student-teacher relationships on childrens’ social and cog- relationships and the trajectory of children’s school out-
nitive development. Educational Psychologist, 38(4), comes through eighth grade. Child Development, 72(2),
207–234. https://doi.org/10.1207/S1532698SEP3804_2 625–638. https://doi.org/10.1111/1467-8624.00301
DiDonato, A. (2014). New directions in social competence Hattie, J. (2009). Visible learning. A synthesis of over 800 meta-
research: Examining developmental trajectories and lan- analyses relating to achievement. New York; Routledge.
guage minority populations [Dissertation for the degree Hukkelberg, S., & Sørlie, M.-A. (in review). The Short Collective
Doctor of Philosophy]. Arizona State University. Teacher Efficacy Scale: Construct and criterion validity.
Eccles, J. S. (1999). The development of children ages 6 to 14. Contemporary Educational Psychology.
The Future Of Children / Center for the Future of Children, Humphrey, N., & Wigelsworth, M. (2012). Modeling the
the David and Lucile Packard Foundation, 9(2), 30–44. factors associated with children’s mental health difficulties
https://doi.org/10.2307/1602703 in primary school: A multilevel study. School Psychology
Eiden, R. D., Colder, C., Edwards, E. P., & Leonard, K. E. Review, 41(3), 326.
(2009). A longitudinal study of social competence among Iarocci, G., Yager, J., & Elfers, T. (2007). What
children of alcoholic and nonalcoholic parents: Role of gene-environment interactions can tell us about social
parental psychopathology, parental warmth, and competence in typical and atypical populations. Brain
self-regulation. Psychology of Addictive Behaviors, 23(1), and Cognition, 65(1), 112–127. https://doi.org/10.1016/j.
36–46. https://doi.org/10.1037/a0014839 bandc.2007.01.008
Eisenberg, N., Fabes, R. A., Murphy, B., Maszk, P., Smith, M., Jung, T. &., & Wickrama, K. A. S. (2008). An introduction to
& Karbon, M. (1995). The role of emotionality and regula- latent class growth analysis and growth mixture modeling.
tion in children’s social functioning: A longitudinal study. Social and Personality Psychology Compass, 2(1), 302–317.
Child Development, 66(5), 1360–1384. https://doi.org/10. https://doi.org/10.1111/j.1751-9004.2007.00054.x
1111/j.1467-8624.1995.tb00940.x Kellam, S. G., Ling, X., Merisca, R., Brown, C. H., &
Elliott, S. N., Gresham, F. M., Freeman, T., & McCloskey, G. Ialongo, N. (1998). The effect of the level of aggression in
(1988). Teacher and observer ratings of children’s social the first grade classroom on the course and malleability of
skills: Validation of the social skills rating scales. Journal of aggressive behavior into middle school. Development and
Psychoeducational Assessment, 6(2), 152–161. https://doi. Psychopathology, 10(2), 165–185. https://doi.org/10.1017/
org/10.1177/073428298800600206 S0954579498001564
Fraley, R. C., & Roberts, R. W. (2005). Patterns of continuity: Kjøbli, J., & Sørlie, M.-A. (2008). School outcomes of a
A dynamic model for conceptualizing the stability of indi- community-wide intervention model aimed at preventing pro-
vidual differences in psychological constructs across the blem behavior. Scandinavian Journal of Psychology, 49(4),
life course. Psychological Review, 112(1), 60–74. https:// 365–375. https://doi.org/10.1111/j.1467-9450.2008.00648.x
doi.org/10.1037/0033-295X.112.1.60 Klasussen, T., & Rasmussen, L.-M. P. (2013). Psychometric
Goddard, R. D. (2001). Collective efficacy: A neglected con- validity and reliability of the Social Skills Improvement
struct in the study of schools and student achievement. System. Rating Scales (SSIS-RS) [Master thesis].
Journal of Educational Psychology, 93(3), 467–476. https:// University of Tromsø, Department of Psychology, Faculty
doi.org/10.1037/0022-0663.93.3.467 of Health Science.
Goddard, R. D., Hoy, W. K., & Woolfolk Hoy, A. (2000). Kokko, K., Tremblay, R. E., Lacourse, E., Nagin, D. S., &
Collective efficacy: Its meaning, measure, and impact on Vitaro, F. (2006). Trajectories of prosocial behavior and
S86 M.-A. SØRLIE ET AL.

physical aggression in middle childhood: Links to adoles- adolescence. Journal of Adolescence, 29(6), 857–889.
cent school dropout and physical violence. Journal of https://doi.org/10.1016/j.adolescence.2006.04.009
Research on Adolescence, 16(3), 403–428. https://doi.org/ OECD. (2015). Skills for social progress: The power of social
10.1111/jora.2006.16.issue-3 and emotional Skills. OECD skills studies.
Koomen, H. M. Y., & Jellesma, F. C. (2015). Can closeness, Ogden, T. (1995). Kompetanse i kontekst. En studie av risiko og
conflict, and dependency be used to characterize students’ kompetanse hos 10 og 13 åringer. Prosjekt Oppvekstnettverk.
perceptions of the affective relationship with their teacher? Rapport 3. Barnevernets Utviklingssenter.
Testing a new child measure in middle childhood. British Ogden, T. (1998). Elevatferd og læringsmiljø. Læreres erfaringer
Journal of Educational Psychology, 85(4), 479–497. https:// med og syn på elevatferd og læringsmiljø i grunnskolen.
doi.org/10.1111/bjep.2015.85.issue-4 Rapport. Kirke-, utdannings- og forskningsdepartementet.
Korpershoek, H., Harms, T., de Boer, H., van Kuijk, M., & Ogden, T. (2003). The validity of teacher ratings in adoles-
Doolaard, S. (2016). A meta-analysis of the effects of class- cent’s social skills. Scandinavian Journal of Educational
room management strategies and classroom management Research, 47(1), 63–74. https://doi.org/10.1080/
programs on students’ academic, behavioral, emotional, 00313830308605
and motivational outcomes. Review of Educational Ogden, T. (2015). Sosial kompetanse og problematferd blant
Research, 86(3), 643–680. https://doi.org/10.3102/ barn og unge. Gyldendal Akademisk.
0034654315626799 Ogden, T., & Hagen, K. A. (2018). Adolescent mental health.
Ladd, G. W., Kochenderfer, B. J., & Coleman, C. C. (1996). Prevention and intervention (2nd ed). Routledge.
Friendship quality as a predictor of young children’s early Parks, F. R., & Kennedy, J. H. (2007). The Impact of race,
school adjustment. Child Development, 67(3), 1103–1118. physical attractiveness, and gender on education majors’
https://doi.org/10.2307/1131882 and teachers’ perceptions of student competence. Journal
Lindberg, E., & Ogden, T. (2001). Elevatferd og læringsmiljø of Black Studies, 37(6), 936–943. https://doi.org/10.1177/
2000. En oppfølgings-undersøkelse av elevatferd og 0021934705285955
læringsmiljø i grunnskolen. Rapport. Læringsmiljøsenteret. Pianta, R. C., Steinberg, M. S., & Rollins, K. B. (1995). The
Malecki, C. K., & Elliot, S. N. (2002). Children’s social beha- first two years of school: Teacher-child relationships and
viors as predictors of academic achievement: Longitudinal deflections in children’s classroom adjustment.
analysis. School Psychology Quarterly, 17(1), 1−23. https:// Development and Psychopathology, 7(2), 295–312. https://
doi.org/10.1521/scpq.17.1.1.19902 doi.org/10.1017/S0954579400006519
Masten, A. S., Desjardins, C. D., McCormick, C. M., Pianta, R. C., & Stuhlman, M. W. (2004). Teacher-child
Kuo, S. I., & Long, J. D. (2010). The significance of child- relationships and children’s success in the first years of
hood competence and problems for adult success in work: school. School Psychology Review, 33(3), 444–458.
A developmental cascade analysis. Development and Renk, K., & Phares, V. (2004). Cross-informant ratings of
Psychopathology, 22(3), 679–694. https://doi.org/10.1017/ social competence in children and adolescents. Clinical
S0954579410000362 Psychology Review, 24(2), 239–254. https://doi.org/10.
Mpofu, E., Thomas, K. R., & Chan, F. (2004). Social compe- 1016/j.cpr.2004.01.004
tence in Zimbabwean multicultural schools: Effects of eth- Rudolph, K. D., Hammen, C., & Burge, K. (1995). Cognitive
nic and gender differences. International Journal of representations of self, family, and peers in school-age
Psychology, 39(3), 169–178. https://doi.org/10.1080/ children: Links with social competence and sociometric
00207590344000385 status. Child Development, 66(5), 1385–1402. https://doi.
Muthén, L. K., & Muthén, B. O. (2007). Mplus user’s guide org/10.1111/j.1467-8624.1995.tb00941.x
(6th ed.). Muthén & Muthén. Sallquist, J. V., Eisenberg, N., Spinrad, T. L., Reiser, M.,
Nantel-Vivier, A., Kokko, K., Caprara, G. V., Pastorelli, C., Hofer, C., Zhou, Q., Liew, J., & Eggum, N. (2009).
Gerbino, M. G., Paciello, M., Côtė, S., Pihl, R. O., Positive and negative emotionality: Trajectories across six
Vitaro, F., & Tremblay, R. E. (2009). Prosocial develop- years and relations with social competence. Emotion, 9(1),
ment from childhood to adolescence: A multi-informant 15–28. https://doi.org/10.1037/a0013970
perspective with Canadian and Italian longitudinal studies. Sameroff, A. J., & Mackenzie, M. J. (2003). Research strategies
Journal of Child Psychology and Psychiatry, 50(5), 590–598. for capturing transactional models of development: The lim-
https://doi.org/10.1111/j.1469-7610.2008.02039.x its of the possible. Development and Psychopathology, 15(3),
Obradović, J., Burt, K. B., & Masten, A. (2010). Testing a dual 613–640. https://doi.org/10.1017/S0954579403000312
cascade model linking competence and symptoms over 20 Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002).
years from childhood to adulthood. Journal of Clinical Experimental and quasi-experimental designs for general-
Child & Adolescent Psychology, 39(1), 90–102. https://doi. ized causal inference. Houghton Mifflin.
org/10.1080/15374410903401120 Sørlie, M.-A., Hagen, K. A., & Ogden, T. (2008). Social
Obradović, J., & Hipwell, A. (2010). Psychopathology and competence and antisocial behavior: Continuity and dis-
social competence during the transition to adolescence: tinctiveness across early adolescence. Journal of Research
The role of family adversity and pubertal development. on Adolescence, 18(1), 121–144. https://doi.org/10.1111/j.
Development and Psychopathology, 22(3), 621–634. 1532-7795.2008.00553.x
https://doi.org/10.1017/S0954579410000325 Sørlie, M.-A., & Nordahl, T. (1998). Problematferd i skolen:
Obradović, J., Van Dulmen, M. H. M., Yates, T. M., Hovedfunn, forklaringer og pedagogiske implikasjoner.
Carlson, E. A., & Egeland, B. (2006). Developmental Rapport 12a/98. Norsk institutt for forskning om oppvekst,
assessment of competence from early childhood to middle velferd og aldring.
 INTERNATIONAL JOURNAL OF SCHOOL & EDUCATIONAL PSYCHOLOGY S87

Sørlie, M.-A., & Ogden, T. (2014). Reducing threats to valid- UNICEF. (2012). Global evaluation of life skills education
ity by design in a nonrandomized experiment of a programmes.
school-wide prevention model. International Journal of Welsh, M., Parke, R. D., Widaman, K., & O’Neil, R. (2001).
School & Educational Psychology,4, 2(4), 235–246. https:// Linkages between children’s social and academic competence:
doi.org/10.1080/21683603.2014.881309 A longitudinal analysis. Journal of School Psychology, 39(6),
Sørlie, M.-A., & Ogden, T. (2015). School-wide positive behavior 463–481. https://doi.org/10.1016/S0022-4405(01)00084-X
support-Norway. Impacts on problem behavior and classroom Welsh, W. N. (2003). Individual and institutional predic-
climate. International Journal of School & Educational tors of school disorder. Youth Violence and Juvenile
Psychology, 3(3), 1–16. https://doi.org/10.1080/21683603.2015 Justice, 1(4), 346–368. https://doi.org/10.1177/
Sørlie, M.-A., & Torsheim, T. (2011). Multilevel analysis of 1541204003255843
the relationship between teacher collective efficacy and Wentzel, K. R., Battle, A., Russell, S. L., & Looney, L. B.
problem behaviour in school. School Effectiveness and (2010). Social supports from teachers and peers as predic-
School Improvement, 22(2), 175–191. https://doi.org/10. tors of academic and social motivation. Contemporary
1080/09243453.2011.563074 Educational Psychology, 35(3), 193–202. https://doi.org/
Spilt, J. L, Vervoort, E, & Verschueren, K. (2018). Teacher- 10.1016/j.cedpsych.2010.03.002
child dependency and teacher sensitivity predict engage- Westiling, E., Andrews, J. A., & Peterson, M. (2012). Gender
ment of children with attachment problems. School differences in pubertal timing, social competence, and
Psychology Quarterly, 33(3), 419–427. https://doi.org/10. cigarette use: A test of the early maturation hypothesis.
1037/spq0000215 Journal of Adolescent Health, 51(2), 150–155. https://doi.
Spivak, A. L., & Farran, D. C. (2016). Predicting first graders’ org/10.1016/j.jadohealth.2011.11.021
social competence from their preschool classroom inter- Zee, M., Koomen, H. M. Y., & Van Der Veen, I. (2013).
personal context. Early Education and Development, 27(6), Student–teacher relationship quality and academic adjust-
735–750. https://doi.org/10.1080/10409289.2016.1138825 ment in upper elementary school: The role of student
Tremblay, R. E., Loeber, R., Gagnon, C., Charlebois, P., personality. Journal of School Psychology, 51(4), 517–533.
Larivée, S., & LeBlanc, M. (1991). Disruptive boys with https://doi.org/10.1016/j.jsp.2013.05.003
stable and unstable high fighting behavior patterns during Zsolnai, A. (2002). Relationship between children’s social
junior elementary-school. Journal of Abnormal Child competence, learning motivation and school achievement.
Psychology, 19(3), 285–300. https://doi.org/10.1007/ Educational Psychology, 22(3), 317–330. https://doi.org/10.
BF00911232 1080/01443410220138548