Faculty of Social and Behavioural Sciences
Graduate School of Child Development and Education
Student–Teacher Relationship Quality From Kindergarten Through Sixth Grade: the Role of Family and Background Factors
Research Master Child Development and Education Master Thesis
Name: A.J. Daniels Supervisor: Dr. M. Zee
Reviewers: Dr. D.L. Roorda and Dr. H.M.Y. Koomen University of Amsterdam, Amsterdam, March 2020
Abstract
The development of student–teacher relationship quality was examined from a Dutch nationally representative sample of 1458 students between kindergarten and sixth grade. The first study aim was to examine trajectories of change in the student–teacher relationship quality (closeness, conflict, and dependency) between kindergarten, third grade, and sixth grade. The second aim was to assess whether proximal factors (i.e., gender, ethnicity, externalizing behavior) and distal factors (i.e., parental involvement in education, home language, maternal years of education, family composition) predict these trajectories of change. Using data from the national Dutch COOL-cohort study, teachers reported on the degree of closeness, conflict, and dependency at each of the three measurement occasions and on parental involvement and externalizing problem behavior at the first measurement
occasion. At the first measurement occasion parents reported on family composition, educational background, home language, and ethnicity. Latent Growth Curve analysis indicated that on average, student–teacher relationship quality improved throughout the elementary school with a linear increase in closeness levels, a quadratic decrease in conflict levels, and a cubic decrease in dependency levels. The results also showed that the boys, students with more externalizing behaviors and minority status developed lower-quality student–teacher relationship in kindergarten and the subsequent elementary school years. Last, some distal factors (parental involvement, maternal years of education) contributed to the development of student–teacher relationship quality over time. The results of this study might be used in future studies promoting equal opportunities for all students to develop high-quality relationships with their teachers.
Keywords: Student–teacher relationships; Elementary school period; Ethnicity; Background factors, Latent growth curve analysis
Student–Teacher Relationship Quality From Kindergarten Through Sixth Grade: the Role of Family and Background Factors
Extensive research has shown that the development of high-quality student–teacher relationships are important for students' social and academic adjustment (Lei, Chui, & Chiu, 2016; Nurmi, 2012; Roorda, Jak, Zee, Oort, & Koomen, 2017; Roorda, Koomen, Spilt, & Oort, 2011). A high-quality relationship, marked by high levels of closeness and low levels of conflict and dependency, enables a secure base for the student to explore its environment (Verschueren & Koomen, 2012). Moreover, high-quality student–teacher relationships may decrease the chance of students developing problems at school, including poor academic achievement and behavior problems (Berry & O’Connor, 2010; Murray & Zvoch, 2011; Maldonado-Carreño & Votruba-Drzal, 2011; Silva et al., 2011). Given the impact of high-quality student–teacher relationships on children’s development, it is important to investigate the potential risk factors that might influence the development of these relationships
throughout the elementary school period.
Prior research has already identified some factors that contribute to the development of student–teacher relationship quality. More specifically, multiple studies on these risk factors have indicated that proximal factors, such as student’s gender, ethnicity, and behavior are associated with the student–teacher relationship quality at a given moment in time (e.g., Lei et al., 2016; Nurmi, 2012). However, less is known about how these student characteristics (proximal factors) predict the trajectories of student–teacher relationship quality over time. A limited number of studies have been conducted examining the development of student– teacher relationship quality over time and they did not always comprise the complete elementary school period (e.g., Jerome, Hamre, & Pianta, 2009; O’Connor, 2010; Pianta & Stuhlman, 2004). Furthermore, compared to the knowledge about proximal factors, even less information is available about how distal factors, such as the home environment or parental
education, impact the development of student–teacher relationship quality throughout
elementary school. Information on how the student–teacher relationship quality is related with distal factors is mainly derived from cross-sectional studies (Murray & Murray, 2004; Wyrick & Rudasill, 2009), but longitudinal studies on the topic are scarce. Additionally, although dependency is considered one of the three dimensions to measure relationship quality, it has often been left out in studies, hence a limited amount of information is available on how dependency develops over time. This is unfortunate because dependency appears to be a stronger predictor for academic adjustment than conflict (Zee, Koomen, & van der Veen, 2013). Thus, it is desirable to extent our knowledge of student–teacher relationship development and investigate how both proximal and distal factors are related to the
development of conflict, closeness and dependency throughout the entire elementary school period.
In order to get a better understanding of these particular relationships, we will first examine initial levels of conflict, closeness, and dependency between teachers and students at kindergarten and investigate general trends growth of conflict, closeness, and dependency from kindergarten through sixth grade. Second, we will examine variations in these growth trajectories based on proximal factors (i.e., gender, ethnicity, and externalizing behavior) and distal factors (i.e., home language, maternal years of education, parental involvement, and family composition). A more thorough understanding of how relationship quality develops over time, could highlight risk and protective factors for teachers and students, and eventually might help teachers to establish high-quality relationships, also with the students who are at risk of developing low-quality relationships.
A Developmental Systems Perspective on Student–Teacher Relationships
Empirical research on the development of student–teacher relationships has been largely inspired by Developmental Systems Theory (Myers & Pianta, 2008; Pianta, Hamre, &
Stuhlman, 2003). This theory is based on the idea that students’ development can be seen as a function of dynamic processes that are embedded in interactions between the student and different contexts (Myers & Pianta, 2008). Developmental systems theory explains that an individual is shaped by the context and can shape its context over time (Lerner, 2002). One of the contexts in which a child develops is the student–teacher relationship.
According to developmental systems theory, student–teacher relationships are formed by external influences from other contexts, characteristics of the student and teacher within the context of the relationship, and interactions between the student and teacher (Meyers & Pianta, 2008).Within the context of the relationship, proximal factors of influence on the student–teacher relationship are the characteristics of the students, including among others, gender, temperament, personality and intelligence (Pianta et al., 2003). More distal factors of influence come from outside of the student–teacher relationship context (Pianta et al., 2003). Examples of factors that contribute to the development of student–teacher relationships are the family environment (Bronfenbrenner, 1986) and parental behaviors (Sameroff, 1989). In this study, students' home language, mothers' years of education, family composition, and parental involvement are examined. Both proximal factors and distal factors form interactions between students and teachers and, subsequently, the quality of student–teacher relationships (Sabol & Pianta, 2012).
The student–teacher relationship quality is often characterized by the three dimensions of closeness, conflict, and dependency (Pianta, 2001; Verschueren & Koomen, 2012).
Closeness refers to the degree of warmth and open communication in the relationship, whereas conflict entails the level of discordance and negativity in the relationship
(Verschueren & Koomen, 2012). Finally, dependency can be defined as students' overreliance on the teacher and the degree of possessiveness in the relationship (Pianta, 2001). Based on developmental systems theory, we will first describe empirical studies on growth trajectories
of student–teacher relationship quality, and thereafter we will look at studies that consider the influence of proximal and distal factors on the development of student–teacher relationship quality.
Growth Trajectories of Closeness, Conflict, and Dependency
Some longitudinal studies have investigated how conflict and closeness and, to a lesser extent dependency, between students and teachers develop across time and found partly contradicting results (Jerome et al., 2009; O’Connor, 2010; Pianta & Stuhlman, 2004). With regard to the development of closeness, a study that assessed teacher-reported closeness between students and teachers in preschool, kindergarten, and first grade found a decrease in closeness over time (Pianta & Stuhlman, 2004). Extending these findings to the upper elementary grades, Jerome et al. (2009) found a decrease in closeness between students and teachers over time, and a faster decrease was measured between third and fifth grade
compared to kindergarten and third grade. Other researchers used relationship trajectories to identify subgroups of students with specific developmental paths of the relationship quality. Also focusing on the entire elementary school period, Bosman, Roorda, van der Veen, and Koomen (2018) showed three different trajectories for teachers-perceived closeness. The most favorable trajectory, high, stable levels of closeness, included the majority of the students, but a small group of students followed high decreasing trajectories of closeness. Another study with pre-kindergarten through fifth-grade students indicated that the largest group, including half of the students, followed stable development across the elementary school, with high levels of closeness (O’Connor, Collins, & Supplee, 2012). Other, smaller groups showed a moderate increase, but also a high decrease and a low, stable level of closeness. Hence, we expected to find a decrease in closeness levels over time.
Considering teacher-reported conflict, Pianta and Stuhlman (2004) found in their longitudinal study a modest decrease in conflict between preschool and first grade in a sample
of 490 children. On the contrary, Jerome et al. (2009) and O’Connor (2010) found an increase in teacher-perceived conflict between kindergarten and fifth grade. The latter studies included multiple measurement occasions from a more extended period, including middle and later elementary school years. Considering the growth trajectories, again, most students were part of the group that had low levels of conflict in their student–teacher relationships (Bosman et al., 2018; O’Connor et al., 2012). However, other trajectories in which teachers reported a small increase in conflict over time and trajectories in which teachers reported a high decrease in conflict over time were also identified. Therefore, we expected to find increase in levels of conflict over time.
The development of dependency over time is less often investigated. Pianta, Steinberg, and Rollins (1995) found in their longitudinal study a stable pattern of dependency between kindergarten and second grade with a sample of 436 students. Unfortunately, this provides limited information about how this relationship dimension develops in subsequent years because a different pattern than the observed one might occur in higher grades. However, using a different study method and including a longer period of time, Bosman et al. (2018) examined growth trajectories of dependency from kindergarten through sixth grade and found other results. The largest group of the students (N = 1155) followed low decreasing levels of dependency, and the other part (N = 145) followed small increasing levels of dependency. Altogether we expected to find small decreasing levels of dependency between kindergarten and sixth grade.
In conclusion, the development of student–teacher relationship quality for the majority of the students is relatively stable and also high in closeness and low in conflict and
dependency. However, longitudinal studies that used several methodological strategies also indicated that a part of the students had less favorable relationships with their subsequent teachers during elementary school (Bosman et al., 2018; Jerome et al., 2009; O’Connor, 2010,
2012; Pianta & Stuhlman, 2004). Both proximal and distal factors might play an important role in the development of relationship quality between students and their teachers in elementary school.
Proximal Predictors of the Development of Student–Teacher Relationship Quality Prior research has shown that some students are more likely to develop a high-quality student–teacher relationship at the start of elementary school than others (Ewing & Taylor, 2009; Jerome et al., 2009; Thijs, Westhof, & Koomen, 2011). The majority of these studies had focused on how students’ characteristics affect the development of relationship quality throughout elementary school (i.e., Bosman et al., 2018; Jerome et al., 2009). In the present study, we concentrate on gender, ethnicity, and externalizing behavior.
First, considering gender, meta-analyses have repeatedly shown that boys tend to have more conflictual and less close relationships with their teachers compared to girls (Baker, 2006; Hamre & Pianta, 2001; Wu, Hughes, & Kwok, 2010). A longitudinal American study, with 878 students from kindergarten through sixth grade, investigated the effect of gender on teachers-perceived closeness and conflict (Jerome et al., 2009). It found that teachers reported a stronger decline in closeness with boys compared to girls, which resulted in an increasing gap of closeness levels over time between the two (Jerome et al., 2009). This is in line with the results of another longitudinal study performed with a large ethnically diverse sample (N = 657) of American first to fifth-grade students (Spilt, Hughes, Wu, & Kwok, 2012). In this study, only boys were represented in high stable trajectories of conflict. No gender effects were found for closeness. Hence, we expected that boys would develop lower-quality student–teacher relationships throughout elementary school than girls.
Second, many studies have indicated that students with externalizing problem behaviors, which refer to ongoing patterns of overactive, impulsive, or aggressive behaviors (Nagin & Tremblay, 1999), generally have less close and more conflictual relationships with
their teachers than those without such problems (Ewing & Taylor, 2009; Henricsson & Rydell, 2004; Jerome, et al., 2009; Lei et al., 2016; Murray & Zvoch, 2011). For instance, Jerome et al. (2009) found in their study that student’s externalizing behavior influenced student–teacher relationship quality at kindergarten, yet no effect on the development of student–teacher relationship quality was found. Furthermore, O’Connor (2010) found that students from first through fifth grade with more behavioral problems had lower relationship quality with their teachers on average. Finally, two large-scale person-centered studies showed that students with externalizing behavior were overrepresented in groups that followed trajectories characterized by an increase in conflict (Bosman et al., 2018; Spilt, Hughes, et al., 2012). Hence, it was expected that students who show more externalizing problem behavior would develop lower-quality student–teacher relationship compared to their peers.
Lastly, with regard to ethnicity, studies have indicated that teachers rate their
relationship quality with students from some ethnic minority groups less positively compared to relationships with ethnic majority students (Hamre & Pianta, 2001; Hughes, Gleason, & Zhang, 2005; Jerome et al., 2009; Spilt & Hughes, 2013; Thijs et al., 2012). In a longitudinal study, teachers experienced more conflict with ethnic minority students over time compared to ethnic majority students, yet ethnicity was not related to closeness (Jerome et al., 2009). Similarly, it was found that African-American children in classrooms with less positive climates had substantially lower-quality relationships than their European-American peers (O’Connor, 2010). Finally, Spilt and Hughes et al. (2012) showed in their person-centered study that African American students were overrepresented in the trajectories that were characterized by increasing levels of conflict, and it appeared that ethnicity was more positively related to conflict than negatively related to closeness. A Dutch study was
in elementary school attending grades 4 through 6 with ethnic majority teachers (Thijs et al., 2012). On average, teachers experienced more conflict and dependency in interactions with the Moroccan-Dutch students. Nonetheless, Bosman et al. (2018) could not replicate the same effect of ethnicity in a Dutch sample. Hence, we expected that students with non-western parents were less likely to develop high-quality student–teacher relationships over time compared to their peers.
Distal Predictors of Student–Teacher Relationship Development
In contrast to the available information from empirical studies about the contribution of proximal predictors on the development of student–teacher relationship quality, less information was available on the role of distal predictors. In the present study, we focused on the language spoken at home, parental education, family composition, and parental
involvement.
First, with regard to student’s home language, it is yet unclear whether speaking a different language at home than at school might be beneficial or a disadvantage for students to develop a high-quality student–teacher relationship. On the one hand, a Belgian study showed that six-year-old bilingual children with a Turkish and Moroccan background scored lower on language proficiency in Dutch (Vanbuel, Boderé, & Jaspaert, 2018). It is argued that students with lower language proficiency may be less able to understand others (Menting, van Lier, & Koot, 2011) and are less able to participate in comprehensive conversations (Spilt, Koomen, & Harrison, 2015). This way, it may be harder for teachers to exchange personal experiences with students that have lower language proficiency, making it harder to establish close relationships. Also, when it is harder for students and teachers to understand each other, miscommunication or even conflicts could arise, which might result in more conflict in the relationship. A study among 133 preschoolers attending programs serving at-risk children supports this line of reasoning by showing that lower language abilities were associated with
more conflict and less closeness in the student–teacher relationship (Justice, Cottone, Mashburn, & Rimm-Kaufman, 2008).
On the other hand, students might benefit from speaking multiple languages. Research with Italian and English bilingual children (n = 77) from three to six years old showed that bilingual children were better able to recognize effective communicative responses, such as speaking the truth or being polite, while observing a conversation (Siegal et al., 2010). Thus, bilingual children with sufficient second language proficiency might have an advantage over their monolingual peers. Teachers might evaluate higher levels of closeness and lower levels of conflict and dependency over time with bilingual children since interactions are more meaningful when the student is able to understand how to respond effectively and politely. Moritz Rudasill, Rimm-Kaufman, Justice, and Pence (2006) found in their cross-sectional study with 99 at-risk students that a lower language complexity was related to more conflict in student–teacher relationship in preschool. Moreover, Spilt et al. (2015) found it their
longitudinal large scale study (N = 4,983) that students with better language skills were more likely to develop closer relationships with their elementary school teachers over time.
Considering the different scenario’s, we do not have a firm hypothesis of the contribution of bilingualism on the development of student–teacher relationship quality.
Second, in some studies, parental education has been considered as one of the indicators of social-economic status, in combination with, for example, parental income. Studies measuring the development of student–teacher relationship quality showed that teachers reported on average a higher level of conflict with their students in sixth grade when they came from low-income families (Rudasill, Reio, Stipanov, & Taylor, 2010). Also, students from families with a lower socio-economic status were more likely to develop lower levels of closeness with their teacher over time (Jerome et al., 2009). In the same study, maternal education did not influence the development of student–teacher relationship quality.
Although a limited amount of evidence is available, we expected that students with less educated mothers would develop lower-quality student–teacher relationships than students with highly educated mothers.
Third, with regard to family composition, a limited amount of information was available because studies on student–teacher relationship quality and family composition are scares. Still, it is theorized that students who have secure child-parent relationships are more socially competent compared to students with insecure relationships (Pianta & Steinberg, 1992). A studies about family context and social well-being supports this theory by showing that students who grew up in intact families tended to have advantages in educational attainment and social well-being (McLanahan & Sandefur, 1994). Similarly, students from single-parent families were more likely to exhibited behavior problems (Gringlas &
Weinraub, 1995). However, a study that was actually focused on relationship quality between teachers and 903 students from kindergarten through fifth grade, which included
questionnaires about actual separation or threatened separation of parents from children during childhood, did not find an effect of these items on the student–teacher relationship quality (Kesner, 2000). In conclusion, there was some evidence to hypothesize that students who grow up in a single-family with non-divorced parents would develop higher-quality student–teacher relationships compared to peers who grew up in multiple families.
Finally, a cross-sectional study among 894 third-grade students showed that parental involvement increased the quality of the interpersonal bonds between teachers and students, and countered relational conflict (Wyrick & Rudasill, 2009). For students from low-income families, more parental involvement was associated with less conflict between the student and its teacher. This finding was repeated in a longitudinal study measuring among other things the amount of parental support at home, the contact between parents and the school, and the quality of the student–teacher relationship of children from birth through fifth grade
(O’Connor, 2010). In general, the student–teacher relationship quality declined, but a less rapid decline in closeness was found between first and fifth-grade students when their parents had more contact with the school. Therefore, we hypothesized that students developed higher-quality relationships with their teachers when their parents were more involved in school. Present Study
The present study was conducted to obtain more insight into the development of student–teacher relationship quality (closeness, conflict, dependency) over time and the influence of proximal and distal factors on the development of student–teacher relationship quality. First, we examined what type of growth rate (linear, quadratic, cubic) could be found for teacher-perceived closeness, conflict, and dependency in the student–teacher relationship from kindergarten through sixth grade. Based on the research discussed above (Bosman et al., 2018; Jerome et al., 2009; O’Connor, 2010; Spilt, Hughes, et al., 2012), we expected a slight decrease in levels of closeness, stable levels of dependency, and a small increase in levels of conflict across time. Second, guided by developmental systems model (Pianta et al., 2003), we examined how proximal (students’ ethnicity, externalizing behavior, and gender) and distal background factors (parental involvement, maternal years of education, home language, family composition) were associated with student–teacher relationship quality in
kindergarten. Furthermore, we examined how these proximal and distal background factors influenced the growth rate of the student–teacher relationship quality. We expected that teachers rated higher levels of closeness and lower levels of conflict and dependency with students who showed low levels of externalizing problem behavior, who had a Dutch
ethnicity, and students who had a more favorable background and family characteristics (i.e., high levels of parental education and parental involvement and no co-parenthood) at
kindergarten. Finally, we expected to find a small increase in conflict between students and teachers between the elementary school years for boys, students who showed externalizing
problem behavior, and students with a non-Dutch ethnicity. Due to a lack of information, no specific hypotheses were formulated for the growth trajectories of dependency.
Method Participants
The present research was conducted by making use of the COOL-cohort study data (i.e., COOL refers to Dutch cohort study Education Careers among students aged 5-18; Driessen, Mulder, Ledoux, Roeleveld, & Van der Veen, 2007). This is a large-scale cohort study conducted in the Netherlands in which students and teachers from kindergarten, 3rd grade, and 6th grade participated. The total sample of COOL5-18 consists of 550 regular Dutch primary schools. To arrive at this number, 400 schools that participated in a previous Dutch cohort study (PRIMA; Driessen, Langen, & Vierke, 2004) were asked to continue
participating in the COOL-cohort study. In the PRIMA-study schools were recruited based on the socio-ethnic composition of the student population, province, and degree of urbanization, so that this sample corresponded to the national average. Additionally, 130 disadvantaged schools were selected to include enough students with a migration background (Driessen et al., 2007).
At the start of the study, the initial sample included 13,563 kindergarten students (Driessen et al., 2007). A part of the students did not participate in all waves due to a couple of reasons. First, 208 schools signed off after the first measurement occasion. Second, some students transferred from the partaking school to another school. Only the students who participated in all three measurement occasions were included in the final sample, since it is preferred to include at least three repeated measures per individual in latent growth curve models to achieve model identification (Curran, Obeidat, & Losardo, 2010).
The final sample consisted of 1458 children and their teachers from 231 classes from 122 schools across the Netherlands. At the start of this study, the students (51.2% girls) were,
on average, 5.6 years old (SD = 0.43). The total sample contained a relatively large number of parents with a migration background compared to national figures (Driessen et al., 2007). Most students (72.4%) had two parents who were born in the Netherlands or another Western country. Information was also available on students’ family composition and socio-economic status. A small part of the participating students (3.2%) grew up in more than one family for a part of the time. Additionally, a small part of the mothers (11.5%) only followed primary education, the majority (44.1%) followed higher secondary education, and 29.7% followed higher vocational or scientific education. Finally, the majority of the mothers spoke Dutch with the participating students (91.9%). No information was available concerning the teachers' gender, age, or ethnicity.
Instruments
The student–teacher relationship quality was measured at each of the three time points. All distal and proximal factors were measured at the first wave. The instruments that were used to measure these factors and the student–teacher relationship quality are discussed below.
Teacher-reported student–teacher relationship quality. The quality of the student– teacher relationship was measured with a short, 15-item form of the authorized Dutch version of the Student–Teacher Relationship Scale (STRS; Koomen, Verschueren, & Pianta, 2007). The questionnaire includes three subscales: Closeness, Conflict, and Dependency. The subscale Closeness (5 items) evaluates the degree to which a teacher experiences affection, warmth, and open communication in the relationship with an individual student, with questions such as "I share a warm, affectionate relationship with this child." The subscale Conflict (5 items) measures the extent to which a teacher assesses the relationship with a student as negative and conflictual, including negative interactions and emotions. An example question is, "This child and I always seem to be struggling." The Dependency subscale (5
items) evaluates the degree of overreliance on the teacher and the possessiveness of the student in the relationship. One of the items was, "This child needs to be continually
confirmed by me." Teachers responded on the STRS-items on a 5-point Likert-type scale (1 = definitely does not apply; 5 = definitely applies). Previous studies have shown that the
psychometric properties of the short form of the STRS are adequate (Zee et al., 2013; Zee & Koomen, 2017). In the present study, the reliability of the three dimensions were satisfactory, with Cronbach's alpha's of .87, .88, .87 for Closeness, .93, .93, .93 for Conflict, and .90, .90, .91 for Dependency at wave 1, 2 and 3, respectively.
Teacher-reported externalizing behavior. To measure students’ externalizing behavior, the Behavior subscale (4 items) was used, derived from a larger questionnaire (13 items) on students’ social-emotional development (Jungbluth, Roeleveld, & Roede, 2001). The Behavior subscale was developed by selecting items from other questionnaires, such as the Teacher Report Form (TRF; Driessen et al., 2007; Ivanova, Achenbach, Rescorla, & Dumenci, 2007; Jungbluth et al., 2001). The Behavior subscale measures teachers’ perception of whether a student behaved rudely or did not follow the rules. One of the items was, "This student is often rude." Teachers were asked to rate students' behavior on a 5-point Likert-type scale (1 = completely untrue; 5 = completely true). Psychometric properties, tested in a
previous cohort study (PRIMA; Driessen et al., 2004), were adequate (Jungbluth et al., 2001). The Cronbach's alpha of this scale was .81 in this study.
Parent-reported ethnicity. Students’ parents could answer the question of where they were born by choosing one of twelve presented options (e.g., Dutch, Moroccan, Turkish, Surinamese). If the parents were born in another country than the twelve given options, they could write down their country of origin. The parents’ answers were categorized into western and non-western countries by the researchers. Students’ ethnicity was dummy coded such that
a 0 represented a student with two parents with a western country of origin and a 1 represented students with one or two parents from a non-western country.
Gender. Gender was dummy coded such that boys were given a value of 0 and girls a value of 1.
Teacher-reported parental involvement. The Parental Involvement scale (3 items) was specifically developed for the PRIMA-cohort study (Driessen et al., 2007). This scale comprises the items "In this family, the parents are actively involved in the school," "In this family learning and curiosity are promoted," and "In this family, the parents support the child in learning." Teachers responded on the parental involvement scale on a 5-point Likert-type scale (1 = definitely not true; 5 = definitely true). Construct validity of the scale was adequate (Driessen et al., 2004), and Cronbach's alpha of this scale was satisfactory (α = .89).
Parent-reported maternal years of education. Mothers had to choose from a list what level of education they had completed. The levels of education were transformed into a continuous score that represented the number of years of education that a mother had
followed. For example, kindergarten was represented with a score of 0, 1-3 years of primary education was represented with a score of 3, and higher vocational or scientific education was represented by a score of 16.
Parent-reported home language. To obtain information about students’ home
language, parents answered the question: "Which language does the child speak the most with its mother?" A dummy variable was created with a 0 representing the Dutch language and a 1 representing a foreign language.
Parent-reported family composition. Parents reported on one item about Family Composition: "Does the child also live in a different family for part of the time, for example, in connection with co-parenthood after a divorce of the parents?" The parents could answer the question with "yes" or "no." A binary variable was created with a 0 representing students
that did not live in a different family and a 1 representing students that did live in a different family for part of the time.
Procedure
Data were collected in three waves (2007-2008, 2010-2011, 2013-2014). In each wave, a coordinator was appointed per region, who was responsible for checking, delivering, and collecting the material and explaining the assessment procedures to teachers in the schools. In January and April of each year, after informed consent was given by the parents, questionnaires were completed by parents and kindergarten, third-grade, and fifth-grade teachers during planned school visits. In each wave, teachers reported on their relationship with their students, and during the first wave only, teachers reported on students’
externalizing behavior and parental involvement. Teachers were also responsible for distributing the questionnaires to the parents. During the first wave, parents completed
questions about their education, their home language, their ethnicity, and family composition. The questionnaires were supplied with separate translations in Dutch, Turkish, Arabic, and English. After completion, the parents returned the completed questionnaires to the teacher in a closed envelope. Table 1 provides an overview of the data collection.
Teachers’ response rate was 92.1%, 93.3%, and 93.6% during the three measurement occasions. The response rate of the parents was approximately 70%. At each wave, there was only a small difference or no difference in response rate between parents of advantageous and disadvantageous social-ethnic background.
Table 1
Overview of Research Variables
Analysis
To analyze the developmental trajectories of closeness, conflict, and dependency across the three measurement occasions, multivariate Latent Growth Curve Analysis (Bollen & Curran, 2006) was performed using structural equation modeling in R version 3.5.2 (R Core Team, 2018). There are several advantages of using Latent Growth Curve Analysis. First, it allows each individual to have its owns growth pattern, which is represented by a unique curve that has different intercepts (initial values at the first measurement occasion) and slopes (growth rates across the measurement occasions). Thus, this method enables us to capture the change in average levels of student–teacher relationship quality and individual differences in change over time (Duncan & Duncan, 2009). Second, latent growth curve analysis permits us to investigate both linear and non-linear growth of the relationship dimensions (Willet & Sayer, 1994).
Modeling procedure. We used a 2-step procedure to create a model for growth trajectories of the relationship dimensions and the contribution of the proximal and distal
Wave Grade N Measures Informant
1 (2007-2008) Kindergarten (2nd year) 1458 Relationship quality: o STRS Proximal predictors: o Gender o Ethnicity o Externalizing behavior Distal predictors: o Home language o Family composition o Maternal years of education o Parental involvement Teacher Teacher/Parent Parent Teacher Parent Parent Parent Teacher 2 (2010-2011) 3 1458 Relationship quality: o STRS Teacher 3 (2013-2014) 6 1458 Relationship quality: o STRS Teacher
factors. We first fitted three separate unconditional growth models for closeness, conflict, and dependency to identify the change in these variables over time. For each model, two latent constructs corresponding to the initial level and the slope of each relationship dimension were defined. The parameters of the three measurement occasions were loaded on the intercepts and slopes as indicators for the two latent constructs. The factor loadings for the intercept were set at 1 for each measurement occasion since the intercept is constant for individuals across time. For testing the growth models, the factor loadings of the slopes of the subsequent time points were fixed to correspond to a linear time scale (0, 1, 2), a quadratic time scale (0, 1, 4), and a cubic time scale (0, 1, 8). After determining the best-fitting model, in the second step, the proximal and distal factors were to the model with the best fitting time scale for the data. The unstandardized scores were used to model change trajectories instead of the
standardized scores to avert problems with finding actual growth rates in the sample (Seltzer, Choi, & Thum, 2003).
Missing data. In the present study, the percentage of missing data was between 0.0% and 4.6%. The Full Information Maximum Likelihood (FIML) function was used for cases with missing data (Enders & Bandolos, 2001). FIML implies likelihood values for missing data points by using available values and parameter estimates from the data (Enders & Bandolos, 2001). The assumption for using FIML was met since the data was missing completely at random, as indicated by Little’s MCAR test, χ2 (85) = 95.719, p = 0.200.
Model goodness-of-fit. To estimate the fit of the models, the root mean square error of approximation (RMSEA; Steiger, 1990) with its 90% confidence interval, and the
comparative fit index (CFI; Bentler, 1990) were used. A model was considered to have an acceptable fit with CFI values over .90 and RMSEA values below .05 (Kline, 2005). The significance of single parameters (i.e., intercepts and regression effects) were tested by Wald z tests. Furthermore, to compare models, chi-square statistics were used, whereby a
significant statistic indicated an improvement or decrease in model fit. To compare models with the same amount of degrees of freedom (i.e., the linear, quadratic, and cubic growth models), a change in fit of ΔCFI > .01 was considered significant (Hu & Bentler, 1999).
Results Descriptive Statistics
Descriptive statistics for all study variables are presented in Table 2. The correlations between the Conflict time points were larger compared to the correlations between the Closeness time points and the correlations between the Dependency time points. The
relationship variables Closeness and Conflict at each time point significantly correlated with all proximal predictors (i.e., Gender, Externalizing Behavior, Ethnicity). Dependency was only positively correlated with Externalizing Behavior. With regard to the correlations between the relationship dimensions and the distal predictors (Parental Involvement, Home Language, Maternal Years of Education, Family Composition), only Parental Involvement correlated with all relationship variables at each wave. The correlations between the relationship dimensions and the distal and proximal factors were mostly in the expected directions. For example, a positive correlation was expected between ethnicity and conflict, since a non-western origin is linked with more conflict in the student–teacher relationship. Finally, with regard to Home Language and Maternal Years of Education, correlations were slightly larger at the later time points compared to the first time point.
With regard to the mean scores, Table 2 shows that, in general, students had high-quality student–teacher relationships across the time points, with high values on Closeness and low values on Conflict and Dependency. Finally, a decrease in means of each relationship dimension was observed across the three time points.
Table 2
Zero-Order Correlations, Means, and Standard Deviations
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. Relationship quality: 1. Closeness1 1.00 2. Closeness2 .23** 1.00 3. Closeness3 .15** .19** 1.00 4. Conflict 1 -.38** -.11** -.07** 1.00 5. Conflict2 -.08** -.30** -.11** .34** 1.00 6. Conflict 3 -.10** -.09** -.35** .24** .31** 1.00 7. Dependency 1 -.17** -.04 -.01 .41** .10** .08** 1.00 8. Dependency 2 .05 .06* -.01 .18** .42** .18** .21** 1.00 9. Dependency 3 .01 .01 -.03 .15** .22** .47** .16** .32** 1.00 Proximal Factors: 10. Gender 1 .20** .18** .14** -.13** -.19** -.17** -.00 .03 -.02 1.00 11. Ethnicity 1 -.11** -.07* -.13** .08** .10** .14** .04 .04 .08 -.02 1.00 12. Ext. Behavior 1 -.24** -.10** -.08** .62** .37** .27** .23** .18** .13** -.11** .07** 1.00 Distal Factors: 13. Involvement 1 .29** .11** .12** -.22** -.12** -.16** -.15** -.06* -.13** .03 -.27** -.22** 1.00 14. Language 1 -.04 -.01 -.10** -.01 .02 .08** .00 .07* .06* .01 .47** .02 -.17** 1.00 15. Mat. Education 1 .05* .02 .11** -.02 -.03 -.06* -.04 -.01 -.08** -.01 -.36** -.05 .27** -.35** 1.00 16. Fam. Composit. 1 -.02 -.06* -.03 .03 .06* .01 -.01 .04 .00 -.02 .03 .01 -.06* -.01 .05 1.00 M 3.85 3.67 3.61 1.70 1.66 1.56 2.15 2.15 1.83 - - 3.69 3.89 - 12.16 - SD 0.59 0.60 0.70 0.73 0.76 0.78 0.74 0.82 0.79 - - 0.76 0.77 - 3.37 - Note. Ext. behavior = externalizing behavior; Involvement = parental involvement; Language = Home language; Mat. education =
Latent Growth Curve Analysis of Closeness
First, the growth rate of Closeness was examined. Table 3 shows that the linear growth model provided the best fit to the data, χ2(1, N = 1458) = 11.370, p = .001; CFI =.93;
RSMEA =.05; SRMR =.026 (see also figure 1). Therefore, the proximal and distal predictors of Closeness were subsequently added to the linear growth model, resulting in a good fitting model χ2(8, N = 1458) = 32.129, p < .001; CFI = .98; RSMEA = .045; RSMR = .020. Compared with the linear model without predictors, the model fit improved significantly (Δχ2(8) = 20.8, p < .01; ΔCFI = .05). Significant variance of both intercept (σ2 = 0.08, SE = 0.02, p < .001) and slope (σ2 = 0.03, SE = 0.01, p = .02) indicated significant individual differences between students in initial levels of Closeness and the change rate in Closeness. The Intercept mean (M = 2.48, SD = 0.11, p < .001) and the Slope mean (M = 0.26 , SD = 0.003, p < .01) were significantly different from zero, indicating that, on average, teachers experienced a linear increase in Closeness over time. Notably, the slope mean (M = -0.02, SD = 0.003, p < .001) of the linear model without the predictors showed that teachers experienced a small linear decrease in Closeness in their student–teacher relationships over time.
The results of the regressions of the final model (Table 4) indicated that Girls, (b = 0.20, z = 7.30, p < .001), students with lower degrees of Externalizing Behavior (b = -0.12, z = 6.57, p < .001), and students with highly involved parents (b = 0.17, z = 9.02, p < .001), had higher levels of Closeness with their teachers at kindergarten. From these three variables, only Externalizing Behavior (b = 0.05, z = 3.30, p = .001) and Parental Involvement (b = -0.07, z = -4.72, p = .001) predicted a slower change rate. Although Maternal Years of Education was not associated with the intercept, it did predict a faster linear change rate across the time points (b = 0.01, z = 2.41, p = .020).
Figure 1. Growth curve of student–teacher Closeness
Latent Growth Curve Analysis of Conflict
With regard to the growth curve analysis of Conflict, the fit indices (Table 3) showed that the quadratic growth model provided the best fit to the data, χ2(1, N = 1458) = 0.015, p = .901; RSMEA = .000; CFI = 1.00 (see also figure 2). Therefore, the predictors were added to the Conflict model with a quadratic growth rate, which resulted in a good fitting model, χ2(8, N = 1458) = 44.577, p < .001; RSMEA = .055; CFI = .98; SRMR = .020. The variance of both intercept (σ2 = 0.06, SE = 0.01, p < .001) and slope (σ2 = 0.02, SE = 0.01, p < .001) were statistically significant, which suggested that there were individual differences between students in initial levels and change rate of Conflict. The Intercept mean (M = 4.06, SD = 0.11, p < .001) and the Slope mean (M = -0.27, SD = 0.05, p < .01) were significantly different from zero. This suggested that, holding all other variables constant, teachers generally experienced an increase in Closeness with their students over time.
The regressions in Table 4 show that the initial levels of Conflict were higher for girls (b = -0.12, z = -4.48, p < .001), students who exhibited higher levels of Externalizing
Behavior (b = 0.52, z = 2.41, p = .01), students with non-western parents (b = 0.08, z = 2.14, p = .03), and students with less involved parents (b = -0.07, z = -3.56, p < .001). Only the proximal factors Gender and Externalizing Behavior predicted the quadratic change rate
0 1 2 3 4 5 kindergarten 3 6 M ea n sco re o f Clo seness Grade
across the time points. Gender predicted a slower change rate in quadratic growth across the timepoints (b = -0.03, z = -2.39, p = .020), but Externalizing Behavior a slower change rate in quadratic growth (b = -0.07, z = 9.39, p < .001).
Figure 2. Growth curve of student–teacher Conflict
Latent Growth Curve Analysis of Dependency
Table 3 shows that the cubic growth model provided the best fit to the data for Dependency, χ2(1, N = 1458) = 2.266, p = .132; RSMEA = .029; CFI = 1.00; SRMR = .012 (see also figure 3). The predictors were jointly added to the cubic growth model of
Dependency, resulting in a good fitting model, χ2(8, N = 1458) = 20.015, p < .001; RSMEA = .015; CFI = .99; SRMR = .012. The intercept (σ2 = 0.12, SE = 0.02, p < .001) and slope variance (σ2 = 0.02, SE = 0.003, p < .001) were significant, indicating that there were individual differences between students in initial levels and change rate. The intercept mean (M = 3.09, SD = 0.13, p < .001) and the slope mean (M = -0.05, SD = 0.02, p = .045) were significantly different from zero, indicating that on average Dependency levels in the student– teacher relationship slightly decreased over time.
The results of the regressions (Table 4) indicated that students with higher levels of Externalizing Behavior (b = -0.21, z = -9.69, p < .001). and students with less involved parents (b = -0.06, z = -2.72, p = .01) had lower initial levels of Dependency. Externalizing
0 1 2 3 4 5 kindergarten 3 6 M ea n sco re o f Co nflict Grade
Behaviour was the only factor that significantly predicted the slope of Dependency (b = -0.01,
z = 3.06, p < .01), indicating that students with higher levels of Externalizing Behavior had a
slower rate of change across the time points.
Figure 3. Growth curve of student–teacher Dependency
Table 3
Fit Indices of Estimated Cross-Lagged Models for Closeness, Conflict, and Dependency
χ2 (df) RMSEA (90% CI) CFI SRMR ΔCFI Model teacher-reported Closeness
Intercept Only Model 16884.34 (8)*** 1.164 (1.149 – 1.178) .00 20.75 n/a
Linear Growth Model 11.370 (1)** .084 (.045 – .131) .93 .026 .93 Quadratic Growth Model 44.948 (1)*** .174 (.133 – .219) .72 .052 .21 Cubic Growth Model 64.524 (1)*** .209 (.167 – .253) .58 .062 .14 Linear Growth Model with Predictors 32.129 (8)*** .045 (.030 – .062) .98 .020 .40
Model teacher-reported Conflict
Intercept Only Model 8502.28 (8)*** .826 (.811 – .840) .00 2.99 n/a
Linear Growth Model 2.827 (1) .035 (.000 – .087) .99 .012 .99 Quadratic Growth Model 0.015 (1) .000 (.000 – .031) 1.00 .001 .01 Cubic Growth Model 0.931 (1) .000 (.000 – .068) 1.00 .007 .00 Quadratic Growth Model with Predictors 44.577 (8)*** .055 (.040 – .072) .98 .023 .02
Model teacher-reported Dependency
Intercept Only Model 9874.43 (8)*** .890 (.875 – .904) .00 4.26 n/a
Linear Growth Model 50.341 (1)*** .184 (.143 – .229) .79 .053 .79 Quadratic Growth Model 10.925 (1)** .083 (.043 – .130) .96 .025 .17 Cubic Growth Model 2.266 (1) .029 (.000 – .082) 1.00 .012 .04 Cubic Growth Model with Predictors 20.015 (8)** .032 (.015 – .050) .99 .012 .01 * p < .05; ** p < .01; *** p < .001. 0 1 2 3 4 5 Kindergarten 3 6 M ea n sco re o f Dependency Grade
Table 4
Summary of Estimates from the Final Models of Closeness, Conflict, and Dependency (N=1458)
Closeness Conflict Dependency
Estimate SE 95% CI Estimate SE 95% CI Estimate SE 95% CI
Fixed effects Intercept 2.477* 0.111 [2.260, 2.694] 4.055* 0.107 [3.845, 4.266] 3.093* 0.13 [2.842,3.343] Slope 0.261* 0.087 [0.090, 0.432] -0.268* 0.046 [-0.358, -0.177] -0.047* 0.02 [-0.092,-0.001] Regressions Intercept Gender 0.201* 0.028 [0.147, 0.255] -0.120* 0.027 [-0.172, -0.068] 0.061 0.032 [-0.002, 0.123] Ext. Behavior -0.122* 0.019 [0.085, 0.158] 0.515* 0.018 [-0.551, -0.480] 0.207* 0.021 [-0.249, -0.165] Ethnicity -0.058 0.036 [0.136, 0.014] 0.075* 0.035 [0.006, 0.144] 0.004 0.042 [-0.078, 0.086] Involvement 0.174* 0.019 [-0.129, 0.211] -0.066* 0.019 [-0.103, -0.030] -0.060* 0.022 [-0.104, -0.017] Home Language 0.051 0.058 [-0.064, 0.165] -0.107 0.056 [-0.218, 0.004] 0.041 0.068 [-0.092, 0.173] Mat. Education -0.007 0.005 [-0.016, 0.002] 0.006 0.004 [-0.003, 0.014] 0.001 0.005 [-0.009, 0.011] Fam. Composition -0.022 0.078 [-0.175, 0.130] 0.120 0.075 [-0.028, 0.268] 0.019 0.090 [-0.158, 0.196] Slope Gender -0.008 0.022 [-0.051, 0.035] -0.027* 0.011 [-0.050, -0.005] -0.009 0.006 [-0.020, 0.003] Ext. Behavior 0.048* 0.015 [-0.077, -0.020] -0.073* 0.008 [0.058, 0.088] -0.012* 0.004 [0.004, 0020] Ethnicity -0.021 0.029 [-0.077, 0.035] 0.023 0.015 [-0.007, 0.052] -0.004 0.004 [-0.012, 0.004] Involvement -0.072* 0.015 [-0.102, -0.042] -0.001 0.008 [-0.017, 0.015] 0.007 0.008 [-0.008, 0.022] Home Language -0.048 0.046 [-0.139, 0.042] 0.039 0.024 [-0.008, 0.087] -0.000 0.012 [-0.025, 0.024] Mat. Education 0.009* 0.004 [0.002, 0.016] -0.001 0.002 [-0.005, 0.002] -0.001 0.001 [-0.003, 0.001] Fam. Composition -0.055 0.062 [-0.176, 0.066] -0.026 0.032 [-0.090, 0.037] -0.007 0.016 [-0.039, 0.025] Note. Ext. Behavior = Externalizing Behavior; Involvement = Parental Involvement; Mat. Education = Maternal Education; Fam. Composition. = Family Composition.
Discussion
Different from previous studies, that mainly focused on proximal factors (Jerome et al., 2009; O’Connor, 2010), the present study also analyzed the contribution of distal factors on the student–teacher relationship development. In addition, whereas other studies only included the relationship dimensions closeness and conflict, the present study included the relationship dimension dependency in order to obtain a complete understanding of student– teacher relationship quality. Moreover, the majority of longitudinal studies on relationship quality have focused on a specific period during elementary school, such as from preschool through first grade (Pianta & Stuhlman, 2004), or from first through fifth grade (O’Connor, 2010). Our study focused on the complete elementary school period from kindergarten through sixth grade. Altogether, the present study addressed two main aims. First, we
examined the development of student–teacher relationship quality from kindergarten through sixth grade. Second, we investigated whether student–teacher relationship quality at
kindergarten and its growth rate across the elementary school years could be predicted by proximal and distal factors.
Trends in Student–Teacher Relationships Across Time
The present study showed that on average, student–teacher relationship quality improved throughout elementary school and that the development of each student–teacher relationship dimension followed a different growth trajectory. More specifically, with regard to the development of closeness, the present study showed that as students reach higher elementary school grades, their teachers experience closer student–teacher relationships with them. Moreover, these average levels of closeness increased throughout elementary school at the same pace. This is inconsistent with previous studies, in which average closeness levels declined stably (O’Connor, 2010; Pianta & Stuhlman, 2004, Spilt et al., 2012) or declined increasingly throughout elementary school (Jerome et al., 2009). It has been theorized that
closeness between students and teachers may decrease across elementary school, since the emphasis within student–teacher relationships shifts from emotional support during kindergarten towards student’s academic work in the higher grades (Jerome et al., 2009; Pianta & Stuhlman, 2004). An explanation for our contradicting results could be that the number of student–teacher interactions decreased over time, yet the teacher-perceived quality of these interactions increased, because it is easier to communicate with older students about personal matters than with younger students. Such interactions may, in the long run, lead to higher levels of closeness. Moreover, it has been theorized that student–teacher closeness becomes more important as students grow older and make the transition to secondary school, because older students tend to be less engaged in higher grades, and have to face more academic challenges (Hamre & Pianta, 2001). A higher degree of closeness in upper elementary school could reflect student’s need for emotional support and motivation. This might suggest that teachers could form a buffer against student’s declining motivation.
With regard to the second relationship dimension, conflict, we found that teachers in the upper elementary school grades experienced less conflict with their students compared to kindergarten teachers. Furthermore, a faster decrease in conflict was found between third grade and sixth grade, compared to the degree of conflict in kindergarten and third grade. Similar to the findings of the closeness dimension, this is not in line with our expectations and some previous studies that mostly found an increase in levels of conflict across time (Jerome et al., 2009; O’Connor, 2010). However, it is consistent with a study whose results indicated a decrease in conflict between kindergarten and first grade (Pianta & Stuhlman, 2004) and the observed decrease in conflict between fourth and sixth grade found by Jerome et al. (2009). Discrepancies between our results and those of previous studies may be due to
methodological reasons. Using growth curve analysis, we specifically investigated individual differences between students’ growth rates in conflict over time. However, other regression
studies (Jerome et al., 2009; O’Connor, 2010) looked into mean differences between measurement moments as if students were part of a homogeneous population.
Finally, with regard to dependency, a cubic trend was found, suggesting decreasing levels of dependency across the elementary school years. More specifically, relatively high levels of dependency were experienced by teachers in kindergarten and third grade, but a relatively large decline in dependency was found between third and sixth grade. This is in line with the assumption that as children transition from elementary to secondary school, students become less dependent on teacher’s support and focus more on interactions with their peers for social support (Hargreaves, 2000; Lynch & Cicchetti, 1997). We found similar results, a fast decline of teacher-perceived dependency in the upper elementary school grades in the Netherlands, in which the students are the same age as secondary school students in the previous studies. A shift in students’ focus of support in upper elementary school is in concordance with our findings of a cubic trend.
Overall, three main findings can be derived from the estimated growth trajectories of student–teacher relationship quality. First, the different growth rates of closeness, conflict and dependency across the elementary period indicate that each dimension seems to tap unique aspects of relationship quality (Verchueren & Koomen, 2012). This implies that the
relationship dimensions are not part of a continuum and should be investigated separately. Second, the three different trends also suggest that different mechanisms underlie at the development of each relationship dimension and that the growth trajectories of the dimensions do not run parallel. This may have implications for interventions designed to increase
student–teacher relationship quality. It may be beneficial or even time-efficient to develop interventions that target specific relationship dimensions at specific grades. For example, students may benefit the most from interventions targeting conflict during kindergarten when teachers experience the highest levels of conflict. Third, our results indicated that, compared
to other studies (Jerome et al., 2009; O’Connor, 2010), students follow a more favorable trajectory of student–teacher relationship quality. It may be assumed that when a
kindergartener has a high-quality relationship with its teacher, it is likely that the relationship quality remains high in upper elementary school.
Proximal Predictors of Student–teacher Relationship Development.
In our study, each of the proximal predictors contributed to the development of student–teacher relationship quality. More specifically, it appeared that boys had less close and more conflictual relationships with their teachers at kindergarten, which is consistent with previous findings (Jerome et al., 2009). With regard to the development of the relationship quality, no differences in the change rate of closeness were found, meaning that boys, who started with less close relationships, ended up with less close relationships at the end of elementary school. It is theorized that girls are more engaged in school (Birch & Ladd, 1997) and that girls are more socialized to seek affiliation from teachers (Spilt, Koomen, et al., 2012), which could lead to closer relationships with their teachers. In contrast, we did find a slower change rate of conflict, meaning that teachers were more likely to experience a faster decrease in closeness for girls compared to boys. An explanation could be that boys are more likely to show externalizing behaviors (Beaman, Wheldall, & Kemp, 2007) and our result also showed that students with externalizing behaviors were less likely to establish high-quality student–teacher relationships over time. It is especially beneficial for this group of students to address their adjustment problems early on in kindergarten to prevent relationship problems in upper elementary years.
Second, considering externalizing behavior, as anticipated we found that kindergarten teachers experienced less closeness and more conflict and dependency in their relationships with students who exhibited higher levels of externalizing behavior. Our results are in line with previous studies that found a negative relationship between externalizing behavior and
student–teacher relationship quality (Jerome et al., 2009; O’Connor, 2010; Pianta & Stuhlman, 2004). With regard to the change rate of the positive relationship dimension, closeness, we found unexpected results. It appeared that students with more externalizing problem behaviors in kindergarten experienced a faster increase in teacher-perceived closeness. Previous longitudinal studies found a negative change rate of externalizing behavior and the development of relationship quality (O’Connor, 2010; Pianta & Stuhlman, 2004). However, using a different statistical analysis, Spilt, Hughes, et al. (2012) examined growth trajectories within a group of first through sixth grade students and found that African American students with more behavior problems experienced a slow increase in closeness levels over time. An explanation for an increase in closeness levels for these specific students might be that that teachers expected more challenging behavior from externalizing students over time (Spilt, Hughes, et al., 2012). Possibly, our results reflected decreases in teachers’ expectations for student–teacher closeness across grades, rather than an absolute increase in closeness.
With regard to the negative relationship dimensions, it appeared that students with externalizing behaviors experienced a faster decrease in teacher-perceived conflict and dependency over time compared to their peers who also followed favorable relationship trajectories in general. These results are in contrast with variable centered studies (Jerome et al., O’Connor, 2010; Pianta & Stuhlman, 2004) that showed increasing levels of conflict over time. However, our results are similar to the findings of Bosman et al. (2019) who used a person-centered study. Bosman et al. found that some students with more externalizing behavior experienced decreasing levels of conflict between kindergarten and sixth grade. It is possible that these students were less prepared for school compared to their peers. Yet, as they grow older and spend more time in school, these students might be able to regulate their behavior (Bosman et al., 2019).
Finally, consistent with our expectations, kindergarten teachers tended to experience more conflict in student–teacher relationships with students whom had a migrant background compared to ethnic majority students, but minority status was not related to the change rate.. Other studies found similar results for students’ ethnicity (Jerome et al., 2009; O’Connor, 2010; Spilt, Hughes, et al., 2012). Both Jerome et al. (2009) and O’Connor (2010) found in their longitudinal studies that African-American children had were more likely to have conflicting relationships with their kindergarten teachers, yet no effect of ethnicity on
relationship quality was found over time. Although minority status was not related to change rate, students with a non-western background continued to have a lower-quality relationships than their peers with a western background in the subsequent years. It is possible that cultural misunderstanding and teacher’s bias towards minority students result in more conflictual interactions (Thijs et al., 2012) which could result in higher levels of teacher experienced conflict. People form specific behaviors and expectations as they grow up in a cultural community. When people from different cultural communities interact while having a weak understanding of each other’s ethnic backgrounds, miscommunication or misinterpretation of behaviors could arise (Saft & Pianta, 2001). When teachers and students continue to have miscommunication a more conflicting student–teacher relationship may develop.
Although other studies found associations between student’s migrant background and closeness (O’Connor, 2010; Spilt, Hughes, et al., 2012), we did not. A reason for these
inconsistencies might be that other studies ( O’Connor, 2010) included minority students with an African-American or Latino-American background and the present study included
minority students with a Turkish or Moroccan background. It is possible that cultural differences between teachers and students with an African-American background are larger compared to teachers and students with a Turkish or Moroccan background. For instance, Thijs et al. (2012) found that Turkish-Dutch students had more close relationships compared
to their Dutch peers, but that Moroccan-Dutch students had more conflictual relationships with their teachers. Considering the fact that we did find associations for conflict but not for closeness, it is possible that similar associations as in the study of Thijs et al. were present in our sample. It might be interesting that future studies are conducted by looking further than western or non-western background into differences between ethnic minority groups. Distal Predictors of Student–teacher Relationship Development.
Some distal factors contributed to de development of student–teacher relationship quality at the start of elementary school and/or the subsequent years. More specifically, kindergarten teachers reported more closeness, less conflict, and less dependency in relationships with students who had more involved parents compared to peers with less involved parents, which was in line with our expectations. Wyrick and Rudasill (2009) found in a cross-sectional study among third graders a similar association. They found that more parental involvement was associated with less conflict for all students and even predicted less conflict for students from low-income families. Similarly, O’Connor (2010) found in a longitudinal study that students whose parents had more contact with the school evidenced less rapid rates of decline in relationship quality. Possibly, parental involvement makes it possible for parents and teachers to exchange information about the student, to adjust learning goals, and to make agreements (Hoover-Dempsey & Sandler, 1995). Further, more parental involvement may help to avoid misunderstandings between parents and teachers (Christenson & Sheridan, 2001). This way, parents may provide information to the teacher that could be used to increase the quality of communication with the student and, eventually, the quality of the relationship with the student (Downer & Myers, 2010). Moreover, it has been argued that students with less involved parents are more likely to distrust others and to behave
disruptively (Brook, Lee, Finch, & Brown, 2012), which in turn could lead to more conflict with their teacher (McGrath & Van Bergen, 2015).
With regard to the development of relationship quality from kindergarten through sixth grade, it appeared that students with more involved parents had more close student– teacher relationships in kindergarten, yet over time their growth rate of teacher-perceived closeness was slower. Thus, differences in closeness levels between students with highly involved parents and less involved parents were more prevalent in kindergarten compared to sixth grade. An explanation could be that students with less involved parents had little support from their parents when they entered kindergarten, but over time learned to use teachers as a source of support (Spilt, Hughes et al., 2012). Unfortunately, with regard to dependency and conflict, the effects remained stable in the subsequent years. This might indicate that it was more difficult for students to overcome their disruptive behavior and clinginess towards their teacher when they had little support from their parents.
Another distal factor, maternal years of education, did not contribute to the amount of closeness between kindergarten students and teachers, yet it did influence the development of closeness in the subsequent years. Teachers experienced a faster increase in levels of
closeness over time with students who had mothers that followed more years of education. This is in line with previous study results showing that aspects of mother-child interaction in a problem-solving task, such as mothers’ positive emotional support and quality of instruction, positively affected students’ adjustment in kindergarten and first grade (Pianta & Harbers, 1996; Pianta, Smith, & Reeve, 1991). An explanation could be that higher maternal education might lead to more warmth and open communication between mother and child, which could set an example for the child. As a result, the child might be more engaged and open in
conversations with their teachers, which in turn could lead to more close relationships. Our results indicated that this was mainly true for upper elementary school. It is possible that teachers expect more open conversations in the higher grades compared to kindergarten.
Contrary to our expectations, family composition did not affect the student–teacher relationship quality at kindergarten or the development of relationship quality over time. It was hypothesized that students who grew up in multiple families were less likely to develop high-quality student–teacher relationships. It is reassuring that this was not the case in the present study, indicating that these students may not be more at risk to develop low-quality student–teacher relationships compared to students who grow up in one family. Please note that only 3.2% of the students in the sample grew up in multiple families and that a minimum of 20% of the subjects in a sample was preferred to examine effect of students’ characteristics on relationship quality (Tabachnick & Fidell, 2007). A sample with a higher percentage of students that grow up in different families might result in other conclusions than the present study.
Similarly, student’s home language was not related to the development of student– teacher relationship quality. We did not set a specific expectation about the effect of home language, because we theorized that speaking multiple languages might be either beneficial or a disadvantage for students to develop a high-quality student–teacher relationship. A part of the students that spoke a different language at home might have had lower language
proficiency because they lacked the possibility to practice Dutch at home. Lower language proficiency could result into lower-quality student teacher relationships (Justice, Cottone, Mashburn, & Rimm-Kaufman, 2008). On the other hand, a student who learns multiple languages might develop better skills to communicate with their teacher (Siegal et al., 2010), what might lead to higher-quality relationships over time. If both groups were present in the sample of this study, both effects could have averaged out the main effect of home language on student–teacher relationship quality. Furthermore, in our sample only 8.1% of the students spoke another language at home than Dutch. A study with a sample that includes more
