Meta-analytic associations between the Student-Teacher Relationship Scale and students’ social competence with peers

Abstract

According to developmental psychologists, more supportive and less conflictual relationships with teachers play a positive role in children’s social behavior with peers both concurrently and in the future. This meta-analysis examined the association between teacher-student relationship quality, as measured by the Student-Teacher Relationship Scale (STRS; Pianta, 2001a), and social competence from early childhood through high school. Based on nearly 30,000 students from 87 studies, the weighted average association between teacher-student relationship quality and social competence with peers was r = .31 (z = .32; 95% CI: .28, 37). Neither age nor length of time between assessments were associated with effect size, suggesting that teacher-student relationships continue to be associated with children’s social competence beyond the early years. Additionally, the STRS total score was the best predictor of social competence, whereas dependency was more weakly associated with social competence. The findings of this study suggest that teacher-student relationship quality as measured by the STRS is an important correlate of both concurrent and future social competence from early childhood to adolescence.

Since the introduction of the Student-Teacher Relationship Scale (STRS; Pianta, 2001a), a great deal of research on the developmental significance of teacher-student relationships has been generated by researchers from around the world. Meta-analyses have demonstrated that affective teacher-student relationships are robustly associated with executive functions, academic achievement, and school engagement across grades (Roorda et al., 2011, 2017; Vandenbroucke et al., 2018). Although dozens of studies measuring the association between the STRS and social competence with peers have now been conducted, no meta-analysis has been published that systematically reviews and synthesizes the results from these various sources. This study fills this gap by presenting a quantitative review of the literature on the predictive significance of the STRS for students’ concurrent and subsequent social skills, peer relationship quality, and social acceptance.

Defining Teacher-Student Relationship Quality and Social Competence with Peers from a Developmental Perspective

The affective patterns of interaction between teachers and students are commonly referred to as teacher-student relationship quality. Teacher-student relationships are normatively formed and appear to vary along similar dimensions as parent-child relationships, at least in early childhood (Howes & Ritchie, 1999). For example, children may seek security from their teachers during times of transition or stress, using them as a safe haven (Koomen & Hoeksma, 2003). Just like parent-child relationships, teacher-student relationships are considered high-quality when they support the child’s exploration of the (school) environment, including both its social and academic challenges (Davis, 2003; Verschueren & Koomen, 2012).

Likewise, the measurement of teacher-student relationships closely parallels that of parent-child relationships (Sabol & Pianta, 2012). Specifically, measures of affective teacher-student relationship quality tend to include assessments of the emotional availability of the teacher, the teacher’s responsiveness to the child’s needs, and the teacher’s consistency in meeting the child’s needs. In turn, children’s responses to the teacher’s cues and use of the teacher as a secure base are typically assessed. The most commonly used measure of teacher-student relationship quality, the STRS (Pianta, 2001a), was informed by attachment theory and standardized measures of parent-child attachment like the Attachment Q-Set (Waters, 1995). The STRS is a questionnaire that was designed to be completed by teachers about individual students in their classroom, though it is occasionally completed by students. The STRS assesses three dimensions of teacher-student relationship quality that parallel attachment behaviors (Verschueren & Koomen, 2012). Teacher-student closeness, or the extent to which the teacher and child share a warm, emotional bond, is akin to the child’s use of the adult as a safe haven. Teacher-student conflict, or the extent to which the teacher and child struggle with one another and have negative interactions, is akin to avoidance. Finally, teacher-student dependency represents resistance, or the child’s failure to use the teacher as a secure base from which to explore the environment. Thus, affective teacher-student relationship quality includes both positive dimensions of relationship quality (i.e., closeness) and negative dimensions of relationship quality (i.e., conflict, dependency). Although these dimensions are theoretically distinct, most samples observe a negative relation between closeness and conflict (e.g., Koomen et al., 2012).

Social competence refers to a broad range of skills, traits, and outcomes that emerge as early as infancy and continue to develop through adulthood (Dodge et al., 1986). Social competence encompasses a variety of more specific constructs (Groh et al., 2014; Rose-Krasnor, 1997), including individual social skills, or the ability to use appropriate behaviors when interacting with others in a variety of environments (e.g., self-regulation in social situations, prosocial behaviors). Peer relationship quality is also subsumed under the broader category of social competence and includes indicators like play behavior with peers and friendship quality. Finally, social competence also encompasses social acceptance at a group level, including popularity with peers or sense of inclusion in the classroom. Social competence with peers in particular is considered to be a salient developmental task for school-age children (Sroufe et al., 2009). Indeed, healthy social development in childhood and adolescence is associated with healthier adjustment and lowered risk of mental illness in adulthood (e.g., externalizing and internalizing symptoms, substance use; Jones et al., 2015); greater educational and professional success (e.g., employment, educational attainment; Collins & van Dulmen, 2006; Gest et al., 2006; A. J. Reynolds et al., 2010); and higher levels of competence in social relationships that normatively emerge later in life, such as those with romantic partners (e.g., relationship satisfaction; Collins & Van Dulmen, 2006). Many researchers have therefore concerned themselves with identifying events, contexts, and systems that might predict or promote positive social development in childhood, with the hope of ultimately reducing the incidence of maladaptation in adulthood (Masten, 2018).

The Significance of Teacher-Student Relationships for Social Development

Various explanations have been proffered for the association between affective components of teacher-student relationship quality and students’ social competence with peers. Attachment theorists suggest that relationships with adult caregivers early in life shape students’ relational models and inform their social behavior for years to come (Bowlby, 1982). As a proximal influence, the teacher-student relationship may play a particularly important role in the development of the student and might even have the power to reshape students’ internal working models (Sabol & Pianta, 2012). That is, teachers who take on caregiver-like roles and behave sensitively toward their students may positively influence students’ relational models and promote healthier social functioning in other relationships, including those with peers (Buyse et al., 2011; Sabol & Pianta, 2012). Others have suggested that teachers can fulfill a student’s need to belong and thus promote their social exploration (Goodenow, 1993; Wentzel, 1997, 1998), help students develop a social identity (Alderman, 2004), and even serve a regulatory function for the development of emotional and behavioral skills (Ahnert et al., 2012; Pianta, 1997; Yowell & Smylie, 1999). It follows, then, that a positive relationship with a teacher, particularly in the early years, has the potential to play a positive role in children’s social behavior with peers both concurrently and in future.

Additionally, behavioral theories suggest that teacher-student relationship quality may influence a child’s social development through the processes of modeling and social referencing. For example, an early experiment by White and Kistner (1992) demonstrated that children who observed teachers providing positive behavioral feedback to students rated those students as more likeable than those who received negative feedback. A follow-up study replicated these results and additionally demonstrated that the effect is particularly pronounced when a teacher provides negative feedback to a well-liked child (White & Jones, 2000). Theorists supporting social learning and referencing models thus argue that positive teacher-student relationships play a role in promoting positive social competence with peers by particularly influencing peer relationship quality and social acceptance (Endedijk et al., 2022). These influences contribute to a student’s social standing in the classroom while they are in a relationship with that teacher but may not continue to have a direct positive influence on social competence over longer periods of time.

In support of these theoretical perspectives, existing evidence suggests that teacher-student relationships as measured by the STRS are associated with social development during the school years (e.g., Birch & Ladd, 1998). A relatively early study by Birch and Ladd (1998) presented results from a sample of 199 Kindergarteners in the United States, finding that students who shared close, independent, and low-conflict relationships with their teachers tended to have higher rates of prosocial behavior and lower rates of asocial behavior in Kindergarten. This association persisted when the students were assessed again one year later (Birch & Ladd, 1998). Similar findings were presented for a sample of nearly 1,000 preschoolers followed from preschool through third grade, where STRS conflict was associated with concurrent social skills and subsequent social skills as rated by both parents and teachers (Skalická et al., 2015).

In contrast, some studies have reported null or even negative associations between the STRS and social competence with peers. In a sample of 44 preschoolers, Spritz and colleagues (2010) observed positive correlations between students’ social skills as well as both their STRS-rated closeness and conflict. Furthermore, in a subsample of 490 children who were followed in the Study of Early Child Care and Youth Development, Pianta and Stuhlman (2004) observed associations in the expected direction between STRS closeness and conflict and observed social competence, but inverted associations with teacher ratings of children’s social competence, both concurrently and in subsequent grades.

Given the wide range of findings, and the limited sample sizes in many individual studies, it is difficult to estimate the significance of affective teacher-student relationships for social development more broadly. A combination of random sampling error, measurement error, and meaningful methodological differences between studies likely have contributed to inconsistencies in findings across studies and samples. Thus, the present meta-analysis seeks to (1) summarize associations between teacher-student relationship quality as measured by the STRS and peer social competence, as well as (2) identify methodological and design factors that may contribute to the inconsistencies in findings.

Previous Meta-Analytic Evidence

A recent meta-analysis conducted by Endedijk and colleagues (2022) documented the association between teacher-student relationship quality and peer relationship quality. The estimated overall association was medium in size (i.e., .20 ≤ r < .30; Funder & Ozer, 2019). The authors additionally reported that effect sizes were generally larger in cross-sectional as opposed to longitudinal studies and when the same rater was used for teacher-student relationship quality and peer relationship quality. No differences were observed by school level in the full sample of studies. These results are informative for the hypotheses of the present study, which builds upon the work of Endedijk and colleagues (2022) by including a broader range of social competence outcomes, including social skills and social acceptance in addition to peer relationship quality. The present study also sought to address inconsistencies in definitions and measurement of teacher-student relationship quality across subfields by focusing only on the STRS, which is the most commonly used measure of teacher-student relationship quality within the developmental literature.

Additionally, Roorda and colleagues (2020) published a meta-analysis focusing on teacher-student dependency (as measured by the STRS in 27 out of 28 studies) and prosocial behavior (which is subsumed under the broader category of social skills). Effect sizes were generally small-to-medium in size (i.e., .10 ≤ r < .30; Funder & Ozer, 2019). The authors compared cross-sectional to longitudinal studies and noted that effects were again larger in cross-sectional studies. Consistent with their previous work (Roorda et al., 2011), the authors noted that use of different informants for teacher-student dependency and student outcomes was associated with slightly smaller effect sizes. Roorda and colleagues’ (2020) work informs some of the hypotheses in the present study, but because of the authors’ limited focus on dependency (as opposed to all subscales from the STRS) and prosocial behavior (as opposed to the full spectrum of indicators of social competence), it does not provide a complete picture of the relation between teacher-student relationships and peer social competence.

The Present Study

Given the theorized link between affective components of teacher-student relationship quality and students’ social development, and the ample research that has been conducted to date on this topic, we conducted a meta-analysis to estimate the correlations between teacher-student relationship quality as measured by the STRS and three major indicators of social competence: social skills, peer relationship quality, and social acceptance. The present meta-analysis makes a significant contribution to the literature on teacher-student relationship quality by providing the first systematic meta-analysis of the associations between all subscales of the STRS and a broad range of social competencies. In addition to the main effects meta-analysis, a number of moderator analyses were conducted in an attempt to account for differences across studies in observed effect sizes. Research questions and associated hypotheses were as follows:

First, how strongly correlated are teacher-student relationship quality, as measured by the STRS, and students’ social competence with peers (including social skills, peer relationship quality, and social acceptance)?

We expected to find a positive correlation between positive aspects (e.g., closeness) of teacher-student relationship quality and each domain of social competence, and a negative correlation between negative aspects (e.g., conflict) of teacher-student relationship quality and each domain of social competence. More specific values could not be predicted because of the lack of previous meta-analyses in this specific domain.

Second, to what extent do associations between teacher-student relationships, as measured by the STRS, and students’ social competence with peers vary by developmental factors (i.e., age and time spanned between assessments)?

We expected that the correlation would be associated with the age at which teacher-student relationships were assessed. Teachers tend to take on fewer caregiver-like roles in the later grades (Bokhorst et al., 2010), thus making them less similar to traditional attachment figures. Furthermore, students tend to have multiple teachers in later grades, reducing their time with any given teacher and thus making it less likely for each dyadic teacher-student relationship to influence behavior. Additionally, we expected that as the time between assessments increased, the associated correlation would decrease. As students age and meet new teachers, the influence of a teacher earlier in life may be overshadowed by that of future teachers and other environmental factors.

Third, does the association between teacher-student relationship quality, as measured by the STRS, and social competence vary with methodological factors?

Given that researcher decisions related to design can greatly influence study results (Landy et al., 2020), we sought to identify methodological factors that could amplify or attenuate the observed effect size.

A. Is reporter associated with effect size?

The STRS can be completed by teachers or students, and social competence can be assessed using a variety of methods. However, given that the majority of studies relied on teacher-report for both variables, the correlation between them may be partially attributed to having a common reporter rather than having a meaningful association between the latent constructs (Endedijk et al., 2022). Therefore, we expected to observe smaller correlations when the methods of measurement or reporters for the two variables were not the same.

B. Does the correlation vary for different subscales of the STRS?

Given the differences across studies in terms of which of the three subscales (i.e., conflict, closeness, and/or dependency) are used, we sought to explore whether all three subscales were equally useful for estimating peer social competence. This question was purely exploratory, and no specific hypotheses were generated.

Method¹

The present report presents a meta-analysis of a subset of papers (i.e., those that used the STRS to assess teacher-student relationship quality) collected as part of a larger comprehensive literature search. The following search strategy was designed to be inclusive of all developmental psychology papers concerning the association between teacher-student relationship quality and social competence. We describe first the comprehensive search strategy and operational definition requirements used for the general search, and then explain the specific operational definition requirements that were used for selecting the subsample of studies used in the present analysis (i.e., those that used the STRS to assess teacher-student relationship quality).

Search Strategy

The meta-analytic corpus was assembled using three primary search strategies: database searching, hand-searching, and forward and backward citation searching. The first and most comprehensive step involved database searching. Literature was collected via database searching on October 29, 2019. Access to databases was provided through the University of Minnesota’s online subscriptions. The primary database for the literature search was Educational Resources Information Center (ERIC) on the EBSCO platform. The search strategy was then translated to additional databases, including PsycINFO (Ovid) and Education Source (EBSCO). Additional searches were conducted in two preprint repositories, PsyArXiv and EdArXiv. These discipline-specific databases were supplemented with a search in the Web of Science Core Collection to target grey literature.

The primary search string was developed using Boolean operators and controlled vocabulary terms available in the ERIC Education database. The full electronic search strategy for ERIC is located in Appendix A of the online supplement. For all databases, the search terms took the general structure of requiring at least one term related to teacher-student relationships and one term related to social competence. Similar search strings were created for each database, using the thesaurus of each database to select applicable terms. No additional limits were placed in the search string, including no limits on year of publication.

Following the database search, manual hand-searches were conducted in several modes. First, conference proceedings from meetings of scholarly societies in the last ten years (i.e., 2010–2020) were examined to search for posters and presentations that may not have yet reached publication. Societies that were targeted included the Society for Research in Child Development (SRCD), Society for Research on Adolescence (SRA), the American Educational Research Association (AERA), and the American Psychological Association (APA). Second, the CVs, Google Scholar pages, and ResearchGate pages of all first authors of papers selected for inclusion were examined for other relevant publications or projects. These authors were also directly contacted, when email addresses were available, to request details of any in-progress or unpublished projects for inclusion in the meta-analysis.

The final step of the literature collection phase included forward- and backward-citation searching. Two coders screened the reference lists and citations of studies identified as eligible in earlier phases of literature collection using Google Scholar and Web of Science. A stopping rule was established before data collection began, such that the citation search was terminated after forward- and backward-searching of 10 studies in a row yielded no unique manuscripts for inclusion in the meta-analysis. To reduce the likelihood of bias in the stopping rule, manuscripts were randomly assigned an order for this citation searching process. The literature search was ended in May 2021.

Inclusion Criteria

Population Requirements

Studies needed to include both teachers and students as part of the study sample. Teachers could not be education students or teachers-in-training. Students could range in age from those enrolled in preschool through Grade 12, or equivalent in international contexts (i.e., approximately between the ages of 3 and 18 years). Therefore, samples focused on college-level students or children enrolled in daycare or other out-of-home care settings were not included. Additionally, samples drawn from online schools or classes were excluded. Samples could be drawn from any country; however, the study needed to be published in English, German, Spanish, or Chinese, as individuals were available to screen manuscripts and assist with effect size extraction in those languages only. No additional limitations were placed on sample characteristics.

Operational Definition Requirements

For the larger systematic review, we included a broad range of teacher-student relationship measures that reflect the attachment tradition of assessment taken by developmental psychologists. Given the theorized importance of attachment-like relationships for the development of social competence, the present meta-analysis focused only on measures of teacher-student relationship quality that assessed dyadic and affective components of the relationship. Measures that attempted to quantify closeness, relatedness, conflict, dependency, emotional support, attachment, or an overall positive relationship were thus included. Conversely, measures that focused on unidirectional qualities (e.g., teacher liking of a student), teacher behaviors in the classroom (e.g., instructional effectiveness), or quantity (e.g., frequency of contact between teacher and child) were not considered measures of teacher-student relationship quality. Measures that reflected the relation between a teacher or teachers and multiple students (e.g., “Students at this school are treated with respect by their teachers”) were also not included. For the present study, we further refined this criteria to include only studies that assessed teacher-student relationship quality using the STRS, in its various forms. The decision to use a narrower operational definition allowed for a clearer synthesis of results and more meaningful interpretations of findings (for a discussion, see Cooper, 2009).

Social competence was defined as an individual’s social skills with peers (e.g., ability to make friends, interpersonal awareness, interpersonal problem solving, prosocial behavior, cooperation), peer relationship quality (e.g., initiation into peer group, play behavior, friendship quality), or social acceptance among peers (e.g., popularity, acceptance, rejection, likability). These constructs could be measured using observation, teacher-report, parent-report, student-report, or peer-report/peer-nomination procedures. Notably, we opted to code each study for these methodological factors rather than completing a critical appraisal of the rigor of each study, following recommendations against the use of appraisal checklists (e.g., B. Cook et al., 2017; Jüni et al., 1999).

Statistical Requirements

At a minimum, studies were required to report a correlation and associated sample size for the relation between the STRS and at least one indicator of social competence in order to be included in the meta-regression. The STRS and the measure of social competence could be assessed either concurrently, or with the STRS measurement preceding the social competence measurement (that is, studies that reported a correlation between prior social competence and subsequent teacher-student relationship quality only were not included). Raw correlations were preferred, but adjusted correlations were included when raw correlations were not available. Sample sizes for specific correlations were preferred, but sample sizes for the measures or the overall study were used when needed.

Publication Requirements

No specific publication modes were required for inclusion in the meta-analysis. Findings could be reported in a peer-reviewed manuscript (e.g., journal article) or in a more informal format (e.g., conference presentation). Unpublished works (e.g., dissertation, preprints) were also eligible for inclusion in order to reduce the potential for publication bias influencing meta-analytic results (Page et al., 2021; Vevea et al., 2009).

Deduplication

Duplicate studies collected from multiple sources were detected using the “Duplicate Items” feature in Zotero. In the case that two items appeared to be the same but represented alternate versions (e.g., a dissertation and published journal article), both records were flagged and retained. Duplicate studies that were determined to be eligible for full-text review were examined further and the study with a larger sample size or more comprehensive set of measures was retained. Manuscripts that reported results from the same sample were eligible for inclusion, but only manuscripts that provided unique information (e.g., different age or different measure) were included.

Screening Procedure

All abstracts were screened by two coders to assess whether the study was likely to meet eligibility criteria. Inclusion decisions were recorded using the Rayyan online tool (Ouzzani et al., 2016). Disagreements were resolved after review and discussion. The full texts of identified manuscripts were then screened, following the same procedure.

Coding System and Interrater Reliability

Studies that met inclusion criteria were coded by the first and second authors for descriptive information about the manuscript and the sample analyzed. Data for the calculation of effect sizes and their respective variances were also included. Data entry was completed for the studies identified as relevant that provided effect size data. Two individuals coded 33% of the included studies. Interrater reliability was assessed with Cohen’s unweighted kappa for nominal variables, Cohen’s weighted kappa for ordinal variables, and correlation coefficients for continuous variables. Kappa values greater than .80 and correlation coefficients greater than .90 reflect good agreement between raters (Koo & M. Li, 2016; McHugh, 2012). Discrepancies between coders were resolved through discussion and review.

Manuscript Variables

Each manuscript was assigned a unique ID. Manuscripts were coded for the year of publication (κ = 1.00), language (κ = 1.00), and type of publication (e.g., thesis/dissertation, journal article, book chapter; κ = 1.00). Additional information was coded for the analysis sample used in each included study. Samples were assigned a separate ID to account for the same sample being included in multiple manuscripts. The country in which the sample lived (κ = 1.00) and the school level(s) (e.g., preschool; κ = .95), were also coded.

Effect Variables

For teacher-student relationship quality, effects were coded for the subscale (i.e., total relationship quality, closeness, conflict, or dependency; κ = 1.00), method of assessment (e.g., teacher-report; κ = 1.00), and grade of assessment (κ = .98). Grade was coded such that preschool = −1, Kindergarten = 0, and Grades 1–12 = 1–12, respectively. Grades were converted to United States equivalents as needed. For studies that aggregated more than one grade of student in their results, the average grade of the sample was used. A continuous indicator of grade was used rather than a dichotomous variable (e.g., elementary vs. secondary school) to reflect the gradual, continuous (rather than categorical) changes that occur from one grade to the next in terms of teacher behavior and social development. For social competence, effects were again coded for measure (e.g., Social Skills Rating System; κ = .93), domain (i.e., social skills, peer relationship quality, or social acceptance; κ = .92), method of assessment (e.g., student-report, peer nomination; κ = .87) and grade of assessment (κ = .98). Correlation values were entered (r = 1.00) along with their corresponding sample size (r = .89). Correlations were transformed as needed so that higher values always indicated more positive teacher-student relationships and higher social competence. That is, correlations related to negative teacher-student relationship subscales (i.e., conflict, dependency) were multiplied by −1. Likewise, correlations related to negative social outcomes (e.g., lack of popularity) were also multiplied by −1. Therefore, positive correlations always represented a positive association between teacher-student relationships and social competence, whereas negative correlations represented a detrimental one.

Statistical Analysis

Calculating Individual Effect Sizes

Although the r statistic represents a standardized effect size estimate, its use in meta-analysis is undesirable because estimates of standard errors can be biased (N. C. Silver & Dunlap, 1987). Therefore, each r value was transformed into Fisher’s z statistic (N. C. Silver & Dunlap, 1987), which allows for interpretation of both the effect size and its variance. Fisher’s z was calculated as:

$$z=0.5\times ln\left[\left(1+r\right)/\left(1-r\right)\right]$$

A variance was calculated for each z statistic using the associated sample size as follows:

$$v_z=1/\left(n-3\right)$$

Treatment of Outliers

Effect sizes more than three standard deviations away from the grand mean were identified. Outliers were winsorized, such that extreme values were replaced with ±3 standard deviations, as appropriate. This method allowed for the retention of extreme values but limited the influence of any one effect size on inferences made (Ghosh & Vogt, 2012).

Assessment of Publication Bias

Effect sizes were visualized using funnel plots. Additionally, symmetry of each funnel plot was tested using Egger’s test (Viechtbauer, 2017), accounting for the multilevel structure of the data. A non-significant value suggests that there is no relation between sample size and effect size.

Meta-Regression

The mean effect size for each of the social competence domains (i.e., social skills, peer relationship quality, and social acceptance) was calculated using a multilevel mixed effects model estimated with the metafor package in R (Viechtbauer, 2010). Robust standard errors were calculated with a sandwich estimator using the clubsandwich package (Pustejovsky & Tipton, 2018), which adjusts confidence interval critical values using the Satterthwaite small-sample correction (Hedges et al., 2010). Importantly, this analytic approach accounts for the dependencies in effect sizes that occur when more than one effect size from each study is used, as was the case in the present dataset.

Need for predictor analyses was assessed by examining the I² statistic, which estimates the extent to which differences in effect sizes can be attributed to within- and between-study heterogeneity (Higgins et al., 2003). Models including additional predictors of effect size value were estimated using the same procedure as that outlined above. To answer each research question, relevant variables were entered in unique sets corresponding to each research question. All predictors were then entered simultaneously into a final model to determine whether any had a unique association with effect size.

Results

The data collection and screening process for the larger systematic review, as well as the subsample used for the present analysis, is represented in Figure 1. Using database searches, 3893 records were identified for screening. Immediately, 432 were removed because they duplicated the information in another record. After abstract and full-text screening in Rayyan by two coders, 251 manuscripts were identified as meeting inclusion criteria. Of those, 145 manuscripts were retained for inclusion in the meta-analysis; the remainder were excluded either because authors did not report correlations and did not respond to a request for data, or because authors reported only correlations that were reported in another manuscript.

Figure 1.

Flowchart Depicting the Data Collection and Screening Process

Note: Flowchart depicts the review process for the broader systematic review until the final step, when studies not using the STRS were removed for the present analysis.

Following database searching, forward- and backward citation searches were conducted on manuscripts identified as meeting inclusion criteria. Reference lists and citations were searched in a randomly assigned order. After 10 manuscripts in a row provided no new sources through forward- or backward-searching, we moved on to the next stage. In total, three additional manuscripts were identified with this technique.

Next, we searched conference proceedings of SRCD, AERA, SRA, and APA from the last decade to identify unpublished works. Thirty-seven posters and talks were identified. Presenting authors were contacted with a request for conference materials and/or the associated publication. Of the 17 conference items subsequently provided to us for screening, 10 met criteria for inclusion, one met criteria but did not include a correlation matrix and thus could not be included, one was associated with a publication already identified in a prior step, and five were deemed not relevant for the present study.

Next, we contacted all corresponding authors who had not already been emailed in a prior step with a request for new or unpublished research items, as well as recommendations for any papers that they believe should be included. We received 38 papers in response. Of these, six met criteria for inclusion, seven met criteria but a correlation matrix could not be obtained, 19 had already been screened in an earlier stage, and six were deemed not relevant for the present study.

Finally, and for the purposes of the present meta-analysis, we retained only those research articles that used the STRS to assess teacher-student relationship quality, resulting in a final sample size of 87 unique manuscripts. A description of the included studies appears in Table 1. All studies included in the meta-analysis are indicated with an asterisk in the reference list.

Table 1.

Summary of Studies Included in the Meta-Analysis.

Study	Sample Country	Sample Size1	STRS Subscale(s)	Grades2	Social Competence Measure(s)	Grades2
Abenavoli & Greenberg, 2014	United States	282	Conflict, closeness	K	Child Behavior Scale (Ladd & Profilet, 1996); Strengths and Difficulties Questionnaire (Goodman, 1997); Social Competence Scale (Conduct Problems Prevention Research Group, 1995)	K
Acar et al., 2020	Turkey	94	Conflict, closeness, dependency	G2	Children’s Behavior Questionnaire (Rothbart et al., 1994); School Social Behavior Scale (Merrell, 1993)	G2
Ansari et al., 2020	United States	794–866	Conflict, closeness	K, G1, G2, G3, G4, G5, G6	Social Skills Rating System (Gresham & Elliott, 1990)	G9
Apavaloaie & Brumariu, 2015	Romania	46	Conflict, closeness	K	Social Competence and Behavior Evaluation (LaFrenière & Dumas, 1996)	K
Arbeau et al., 2010	Canada	169	Conflict, closeness, dependency	G1	Child Behavior Scale (Ladd & Profilet, 1996)	G1
Aslan & Boz, 2019a	Turkey	211	Conflict, closeness	PreK	School Social Behavior Scale (Merrell, 1993)	PreK
Aslan & Boz, 2019b	Turkey	211	Conflict, closeness	PreK	Preschool Play Behavior Scale	PreK
Aslan & Boz, 2019c	Turkey	211	Conflict, closeness	PreK	Social Competence and Behavior Evaluation (LaFrenière & Dumas, 1996)	PreK
Bakkaloğlu et al., 2019	Turkey	94	Total score	PreK	Peer nominations; Preschool and Kindergarten Behavior Scale (Merrell, 1994)	PreK
Barbarin et al., 2013	United States	335	Closeness	G1	Social Competence Scale (Barbarin, 2007)	G1
Bierman et al., 2017	United States	461	Conflict, closeness	G2	Child Behavior Scale (Ladd & Profilet, 1996); Friendship Questionnaire (Bierman & McCauley, 1987); Perceived Competence Scale for Children (Harter, 1982); Social Competence Scale (Conduct Problems Prevention Research Group, 1995)	G2
Birch & Ladd, 1998	United States	199	Conflict, closeness, dependency	K	Child Behavior Scale (Ladd & Profilet, 1996)	K
Birch, 2002	United States	343	Conflict, closeness, dependency	G3	Peer nominations	G3
Blacher et al., 2009	United States	98	Conflict, closeness	G1, G2, G3	Social Skills Rating System (Gresham & Elliott, 1990)	G1, G2, G3
Breeman et al., 2015	Netherlands	340	Conflict, closeness	G5	Peer nominations	G5
Buhs et al., 2018	United States	1032	Conflict, closeness	G3, G5	Quality of Child’s Friendship Scale (Clark & Ladd, 2000)	G3, G5
Casillas, 2010	United States	135–145	Closeness	G4	Peer nominations	G4
Coplan et al., 2017	China	1275	Total score	G5	Peer nominations	G5
Coulombe & Yates, 2018	United States	215	Closeness	K	Snack sharing task (O’Connor et al., 1979)	K
DeMaioribus, 2010	United States	854	Total score	PreK	Friends or Foes; Playmate Questionnaire (Vandell, 1995)	G1
Demirtaş-Zorbaz & Ergene, 2019	Turkey	517	Conflict, closeness	G1	Child Behavior Scale (Ladd & Profilet, 1996)	G1
Eisenhower et al., 2007	United States	140	Total score	G1	Social Skills Rating System (Gresham & Elliott, 1990)	G1
Ferreira et al., 2016	Portugal	168	Closeness	PreK	Strengths and Difficulties Questionnaire (Goodman, 1997)	PreK
Fisher et al., 2016	United States	209	Total score	G3	Behavioral Assessment System for Children (C. R. Reynolds & Kamphaus, 1992)	G3
Fujii, 2015	United States	22	Conflict, closeness	G2	Social Skills Improvement System (Gresham & Elliott, 2008); Social interaction observation (Bauminger, 2002)	G2
Garner et al., 2014	United States	145	Conflict, closeness, dependency	PreK	Peer Victimization Measure (Crick et al., 1999)	PreK
Gower et al., 2014	United States	179–181	Conflict, closeness	K	Preschool Social Behavior Scale (Crick et al., 1997)	K
Griggs et al., 2009	United States	44	Conflict, closeness, dependency	PreK	Penn Interactive Peer Play Scale (Fantuzzo et al., 1998)	PreK
Harrison et al., 2007	Australia	125	Conflict, closeness	K	Teacher-Child Rating Scale (Hightower et al., 1989)	K
Hernández et al., 2017	United States	301	Conflict, closeness	K	MacArthur Health and Behavior Questionnaire (Armstrong & Goldstein, 2003); Peer nominations	K
Hosan & Hoglund, 2017	Canada	461	Conflict, closeness	G2	Friendship Quality Questionnaire (Parker & Asher, 1993)	G2
Howes et al., 1998	United States	55	Conflict, closeness	G4	Friendship Quality Scale (Bukowski et al., 1994)	G4
Howes et al., 2011	United States	747	Conflict, closeness	PreK	Revised Peer Play Scale (Howes & Matheson, 1992)	PreK
Howes, 2000	United States	307	Conflict, closeness	G2	Teacher Assessment of Social Behavior Questionnaire (Cassidy & Asher, 1992)	G2
Hughes et al., 2014	Canada	157–192	Conflict, closeness, dependency	K	Child Behavior Scale (Ladd & Profilet, 1996); Play Observation Scale (Rubin, 2001); Strengths and Difficulties Questionnaire (Goodman, 2001)	K
Iruka et al., 2010	United States	95	Conflict, closeness	K, G1, G3, G4, G5	Social Skills Rating System (Gresham & Elliott, 1990)	K, G1, G3, G4, G5
Kim & Cappella, 2016	United States	111	Conflict, closeness	G3	Peer nominations	G3
Kopans, 2001	United States	50	Conflict, closeness, dependency	G3	Social Skills Rating System (Gresham & Elliott, 1990)	G3
Labella et al., 2019	United States	245	Total score	K	Health and Behavior Questionnaire (Armstrong & Goldstein, 2003)	K
W. Li et al., 2014	China	529	Conflict, closeness, dependency	G5	Children’s Loneliness Scale (Asher et al., 1984)	G5
Y. Li et al., 2015	China	543	Conflict, closeness	K	Social Skills Rating System (Gresham & Elliott, 1990)	K
Lin et al., 2016	China	256	Total score	G4	Children’s Loneliness Scale (Asher et al., 1984)	G4
Locke, 2011	United States	42	Total score	G1	Social Skills Rating System (Gresham & Elliott, 1990); Social Responsiveness Scale (Constantino & Gruber, 2005); Social Skills Q-Sort	G1
Mackintosh & McCoy, 2019	United States	3485	Conflict, closeness	PreK	Child Observation Record (High/Scope Press & High/Scope Educational Research Foundation, 2003)	PreK
Madill et al., 2014	United States	628	Closeness	G3	Peer nominations	G3
Magelinskaitė-Legkauskienė et al., 2017	Lithuania	403	Conflict, closeness	G2, G3	Elementary School Social Competence Scale (Magelinskaitė, 2014)	G2, G3
Mathur, 1999	United States	188	Conflict, closeness, dependency	G3	Social Skills Rating System (Gresham & Elliott, 1990)	G3
McIntyre et al., 2006	United States	67	Total score	PreK	Social Skills Rating System (Gresham & Elliott, 1990)	PreK
Mohamed, 2018	Oman	160	Conflict, closeness, dependency	PreK	Social Skills Rating System (Gresham & Elliott, 1990)	PreK
Mousouli, 2007	United States	442	Total score	G3, G4	Behavior Assessment System for Children (C. R. Reynolds & Kamphaus, 1998)	G3, G4
Myers & Morris, 2009	United States	140	Conflict, closeness	PreK	Strengths and Difficulties Questionnaire (Goodman, 1997)	PreK
Olivier et al., 2018	Canada	582	Conflict, closeness	G6	Strengths and Difficulties Questionnaire (Goodman, 2005)	G6
Palermo et al., 2007	United States	95	Conflict, closeness, dependency	PreK	Child Behavior Scale (Ladd & Profilet, 1996)	PreK
Peisner-Feinberg et al., 2001	United States	345–733	Closeness	PreK, K, G2	Classroom Behavior Inventory (Schaefer et al., 1978)	PreK, K, G2
Pianta & Stuhlman, 2004	United States	490	Conflict, closeness	PreK	California Preschool Social Competency Scale (Levine et al., 1970); Observational Record of the Caregiving Environment (NICHD Early Child Care Research Network, 1996, 2002)	PreK
Sabol et al., 2018	United States	211	Conflict, closeness	PreK	inCLASS (Downer et al., 2011)	PreK
Saral, 2020	Turkey	127	Conflict, closeness	PreK	Social Competence and Behavior Evaluation (LaFrenière & Dumas, 1996)	PreK
Serdiouk et al., 2016	United States	1568–1634	Closeness	G3	Peer nominations	G3
Serpell & Mashburn, 2012	United States	1939–2644	Conflict, closeness	PreK, K	Teacher-Child Rating Scale (Hightower et al., 1986)	PreK, K
Sette et al., 2013	Italy	88	Conflict, closeness, dependency	K	Peer nominations; Social Competence and Behavior Evaluation (LaFrenière & Dumas, 1996)	K
Sette et al., 2014	Italy	129	Conflict, closeness, dependency	K	Child Behavior Questionnaire (Putnam & Rothbart, 2006); Peer nominations; Social behaviors and social reputation (Caprara & Pastorelli, 1993); Social Competence and Behavior Evaluation (LaFrenière & Dumas, 1996)	K
Sette et al., 2018	Italy	368	Conflict, closeness, dependency	PreK	Social Competence and Behavior Evaluation (LaFrenière & Dumas, 1996)	PreK
Sette et al., 2019	Italy	131	Closeness	K	Child Behavior Questionnaire (Putnam & Rothbart, 2006); Child Social Preference Scale (Coplan et al., 2004)	K
Sette et al., 2021	Italy	212	Closeness, dependency	PreK	Child Social Preference Scale (Coplan et al., 2004); Preschool Play Behavior Scale (Coplan & Rubin, 1998)	PreK
Shin & Kim, 2008	South Korea	297	Conflict, closeness	K	Social Competence and Behavior Evaluation (LaFrenière & Dumas, 1996)	K
R. B. Silver et al., 2010	United States	241	Conflict, closeness	K	Child Adaptive Behavior Inventory (Cowan et al., 1995)	K
Skalická et al., 2015	Norway	981	Conflict	PreK, G1, G3	Social Skills Rating System (Gresham & Elliott, 1990)	PreK, G1, G3
Spilt et al., 2015	Australia	4707	Closeness	PreK, G1, G3	Strengths and Difficulties Questionnaire (Goodman, 1997)	PreK, G1, G3
Spritz et al., 2010	United States	44	Conflict, closeness	PreK	Social Skills Rating System (Gresham & Elliott, 1990)	PreK
Sucuoğlu et al., 2019	Turkey	117	Total score	PreK	Preschool and Kindergarten Behavior Scale (Merrell, 1996)	PreK
Swanson, 2012	United States	291	Total score	G2	Perceived Competence Scale for Children (Harter, 1982)	G2
Thijs & Koomen, 2009	Netherlands	131	Conflict, closeness, dependency	G1	Perceived social problems	G1
Troop-Gordon & Kopp, 2011	United States	410	Conflict, closeness, dependency	G5	Peer nominations	G5
Valeski, 2000	United States	267–330	Conflict, closeness	K, G2, G3	Child Behavior Scale (Ladd & Profilet, 1996)	K, G2, G3
Valiente et al., 2008	United States	264	Total score	G4	Perceived Competence Scale for Children (Harter, 1982)	G4
Varghese et al., 2019	United States	503	Conflict, closeness	K	Strengths and Difficulties Questionnaire (Goodman, 2001)	K
Verschueren et al., 2012	Belgium	94	Total score	G1	Peer nominations	G1
Vitaro et al., 2012	Canada	450–580	Conflict	K, G1	Friendship Features Interview for Young Children (Ladd et al., 1996); Peer nominations	K, G1
Waajid, 2005	United States	58	Conflict, closeness, dependency	PreK	Social Competence and Behavior Evaluation (LaFrenière & Dumas, 1996)	PreK
Wang et al., 2016	Australia	2857	Conflict	G1	Strengths and Difficulties Questionnaire (Goodman, 1997)	G1
Wang et al., 2018	Australia	1442–2890	Conflict, closeness	PreK, G1, G3, G5	Strengths and Difficulties Questionnaire (Goodman, 1997)	G7
Wildenger & McIntyre, 2012	United States	86	Total score	K	Social Skills Rating System (Gresham & Elliott, 1990)	K
Wilson et al., 2016	United States	195	Closeness, total score	G3	Interpersonal Competence Scale (Cairns et al., 1995); Peer nominations	G3
Zee & Koomen, 2017	Netherlands	510	Conflict, closeness, dependency	G5	Peer nominations	G5
Zeedyk et al., 2016	United States	107	Conflict, closeness, dependency	K	Social Competence and Behavior Evaluation (LaFrenière & Dumas, 1996); Social Skills Improvement System (Gresham & Elliott, 2008)	K
Zhang et al., 2009	China	325	Total score	PreK	Early School Behavior Rating Scale (Caldwell & Pianta, 1991)	PreK
Zhang, 2011	China	443	Total score	PreK	Early School Behavior Rating Scale (Caldwell & Pianta, 1991)	PreK

Note. STRS = Student-Teacher Relationship Scale. PreK = Pre-Kindergarten. K = Kindergarten. G1 = Grade 1. G2 = Grade 2. G3 = Grade 3. G4 = Grade 4. G5 = Grade 5. G6 = Grade 6. G7 = Grade 7. G9 = Grade 9.

¹When sample sizes for individual effects were provided, the range of sample sizes is listed.

²When a study included students from multiple grades, the average grade is listed.

Prior to data entry, we identified the sample associated with each included item. Manuscripts were assigned ID numbers that reflect dependencies in sample across manuscripts. Effect sizes that were reported in more than one research item were only entered once.

Diagnostic Analysis

Missing Data

Percent missing data was calculated for each predictor variable in the model within the included 87 studies. No missing data were present at the manuscript or effect size levels.

Outliers

Examination of density plots of effect sizes suggested that they were normally distributed. Effect sizes more than three standard deviations away from the grand mean were identified. Two large effect sizes (Apavaloaie & Brumariu, 2015; Locke, 2011) and one small effect size (Iruka et al., 2010) were identified as outliers. All three identified outliers came from unusually small samples (i.e., less than 100; Apavaloaie & Brumariu, 2015; Iruka et al., 2010; Locke, 2011), likely explaining the particularly strong effects. In each case, these values were replaced with the value of ±3 standard deviations. This method allowed for the retention of these outliers in the samples while also limiting their influence on the estimated variance and average effect size.

Publication Bias

Funnel plots for the full data set, as well as each outcome domain, are shown in Figure 2. A visual inspection of the plots did not suggest that the current sample is affected by the “file drawer problem,” in that there did not appear to be studies missing from any of the bottom left quadrants. Egger’s test of symmetry, modified for the multilevel structure of the data, also suggested that there was no association between sample size and effect size for the full sample of effects (b = 0.00, p = .226), social skills (b = 0.00, p = .145), peer relationship quality (b = 0.00, p = .356), or social acceptance (b = 0.00, p = .383). Taken together, these results suggest that the sample is not significantly affected by publication bias and can be considered a random sample of the population of effect sizes.

Figure 2.

Funnel Plots of Effect Sizes Aggregated at the Manuscript Level (Top) and Unaggregated (Bottom)

Note: Vertical lines represent the unweighted average effect size. Color version of figure available online only.

Descriptive Analysis

The included sample represented work from 76 different first authors reporting on 29,318 participants recruited from 14 different countries. The most common country of origin was the United States (k = 49), followed by Turkey (k = 8), China (k = 6), Canada (k = 5), Italy (k = 5), Australia (k = 4), and Netherlands (k = 3). Other countries represented were Belgium, Lithuania, Norway, Oman, Portugal, Romania, South Korea (k = 1 each). Two manuscripts were published in Chinese; the rest were in English. All appeared between 1998 and 2021 (Median = 2014, Mode = 2019). Twelve items were theses or doctoral dissertations, six were conference papers or posters, and the remainder appeared in peer-reviewed journals. The majority of the studies included participants of preschool age (k = 47), elementary school age (k = 97), and/or middle school age (k = 43).

The majority of papers reported correlations between social competence indicators and specific subscales of the STRS (i.e., conflict, closeness, and dependency). When both a total score and subscale scores were reported in a manuscript, only subscale scores were used to prevent redundancy in the data and to allow for comparisons across subscales. However, 17 studies reported only the total score and no subscale scores; therefore, correlations with total relationship quality were included in the meta-regression for these research items. Additionally, one study used a student report of closeness and a teacher report of overall relationship quality (Wilson et al., 2016); both scales were included in this case because they represented different reporter viewpoints. All included studies but one (Madill et al., 2014) included teachers as the reporter on the STRS. Madill and colleagues (2014) relied on students as the sole reporters on the STRS. Four other studies (Lin et al., 2016; Swanson, 2012; Valiente et al., 2008; Wilson et al., 2016) used student reports in addition to teacher reports on the STRS. The most common grade of assessment of teacher-student relationships was preschool (k = 29), followed by Kindergarten (k = 24). Studies were identified for all grades from preschool through Grade 6.

Assessments of social competence were varied. The most commonly used measure was the Social Skills Rating System or Social Skills Improvement System (k = 15; Gresham & Elliott, 1990, 2008), followed by the Strengths and Difficulties Questionnaire (k = 9; Goodman, 1997), the Social Competence and Behavior Evaluation (k = 9; LaFrenière & Dumas, 1996), and the Child Behavior Scale (k = 8; Ladd & Profilet, 1996). The most common method of assessing social competence was teacher-report (k = 62), followed by parent-report (k = 17), peer-report or peer nominations (k = 16), student-report (k = 6), and observational measures (k = 6). Additionally, four studies averaged together reports from teachers and parents on the same measure (Ferreira et al., 2016; Spilt et al., 2015; Swanson, 2012; Zhang, 2011). Like teacher-student relationship quality, the most frequent grade of assessment for social competence was preschool (k = 27), followed by Kindergarten (k = 23). All grades from preschool through Grade 7 as well as Grade 9 were represented. Most studies included a cross-sectional (i.e., concurrent) correlation between the STRS and social competence (k = 83). However, a minority also included longitudinal correlations between a prior STRS score and a subsequent social competence indicator (k = 17). The amount of time spanned between assessments ranged from zero (i.e., concurrent) to nine years (Ansari et al., 2020).

Meta-Regression

In the full sample, the average effect size for the association between teacher-student relationship quality, as measured by the STRS, and social competence with peers was estimated to be z = .32 (p < .001, 95% CI: .28, .37) after accounting for dependencies created by having multiple effect sizes from each study. This effect size is equivalent to an average correlation of r = .31. A large amount of heterogeneity was present in effect sizes (I² = 96.75%; SD = 0.50), as shown in Figure 2. The average effect size for social skills was estimated to be z = .38 (p < .001; 95% CI: .32, .43; r = .36; I² = 96.56%; SD = 0.45). The average effect size for peer relationship quality was z = .29 (p < .001; 95% CI: .19, .39; r = .28; I² = 97.65%; SD = 0.46). Finally, the average effect size for social acceptance was z = .28 (p < .001; 95% CI: .21, .35; r = .28; I² = 93.53%; SD = 0.38). The high proportion of unexplained variance for all three outcome variables suggested that additional predictors were needed in the models to explain differences in effect sizes within and across studies.

Timing

The first set of predictor analyses examined the association between developmental factors and effect size (Table 2, Step 2). A variable representing the approximate amount of time between assessments was created by subtracting the grade of STRS assessment from the grade of social competence assessment. Effects that were longitudinal but occurring in the same grade (e.g., fall and spring assessments) were coded as 0.5. As shown in Table 2, there was no association between the amount of time spanned between assessments and effect size. Furthermore, as shown in Table 3, Step 2, the amount of time spanned between assessments was not associated with effect sizes for social skills, peer relationship quality, or social acceptance. Contrary to our hypothesis, the grade in which the STRS was administered was also not associated with effect size for any of the outcome domains (Table 2, Step 2; Table 3, Step 2), suggesting that the STRS had a consistent association with each indicator of social competence across the grades. These effects were consistent in the final versions of the models that included all predictor variables simultaneously (Table 2, Step 4; Table 3, Step 4).

Table 2.

Stepwise Meta-Regression Results for the Association Between the STRS and Social Competence

	b	SE	95% CI
Step 1
Intercept	0.32	0.02	0.28, 0.37
Step 2
Intercept	0.34	0.03	0.29, 0.39
Time between assessments	−0.02	< 0.01	−0.05, 0.01
Grade of STRS assessment	−0.01	0.01	−0.02, 0.01
Step 3
Intercept	0.53	0.06	0.39, 0.66
Method concordance1	−0.20	0.04	−0.29, −0.11
Conflict2	−0.12	0.06	−0.24, 0.00
Closeness2	−0.14	0.05	−0.26, −0.02
Dependency2	−0.27	0.06	−0.40, −0.14
Step 4
Intercept	0.52	0.07	0.38, 0.67
Time between assessments	−0.02	0.01	−0.10, 0.07
Grade of STRS assessment	0.00	0.01	−0.02, 0.02
Method concordance	−0.19	0.05	−0.30, −0.08
Conflict	−0.12	0.06	−0.24, 0.01
Closeness	−0.14	0.06	−0.26, −0.02
Dependency	−0.27	0.06	−0.40, −0.14

Note. b = unstandardized beta coefficient. SE = standard error. CI = confidence interval. STRS = Student-Teacher Relationship Scale. k = 461.

¹Different reporters/methods = 1, same reporter/method for both measures = 0.

²Total STRS score is presented as the comparison group. Conflict was associated with a significantly larger effect than dependency (b = −0.15, SE = 0.03, 95% CI: −0.22, −0.08). Closeness was associated with a significantly larger effect than dependency (b = −0.13, SE = 0.04, 95% CI: −0.21, −0.05). There was no significant difference between conflict and closeness (b = −0.02, SE = 0.02, 95% CI: −0.06, 0.02).

Table 3.

Stepwise Meta−Regression Results for the Association Between the STRS and Each Domain of Social Competence

	Social Skills1				Peer Relationship Quality2				Social Acceptance3
	b	SE	95% CI		b	SE	95% CI			b	SE	95% CI
Step 1
Intercept	0.38	0.03	0.32, 0.43		0.29	0.05	0.19, 0.39		0.28		0.03	0.21, 0.35
Step 2
Intercept	0.38	0.03	0.32, 0.44		0.29	0.05	0.19, 0.39		0.33		0.06	0.19, 0.46
Time between assessments	−0.02	0.01	−0.16, 0.12		−0.01	0.01	−0.15, 0.13		−0.04		0.02	−0.10, 0.02
Grade of STRS assessment	0.00	0.01	−0.03, 0.03		0.00	0.01	−0.06, 0.07		−0.02		0.02	−0.06, 0.02
Step 3
Intercept	0.61	0.06	0.48, 0.75		0.07	0.19	−0.46, 0.61		0.45		0.09	0.25, 0.65
Method concordance4	−0.23	0.07	−0.38, −0.09		−0.04	0.14	−0.82, 0.75		−0.20		0.09	−0.42, 0.02
Conflict5	−0.21	0.06	−0.34, −0.08		0.30	0.16	−0.31, 0.90		−0.02		0.07	−0.20, 0.16
Closeness5	−0.17	0.06	−0.30, −0.05		0.20	0.11	−0.45, 0.86		−0.08		0.07	−0.27, 0.10
Dependency5	−0.37	0.07	−0.52, −0.23		0.13	0.15	−0.27, 0.53		−0.15		0.09	−0.35, 0.05
Step 4
Intercept	0.61	0.06	0.48, 0.74		0.23	0.16	−0.24, 0.70		0.45		0.09	0.25, 0.66
Time between assessments	−0.01	0.01	−0.12, 0.11		−0.04	0.01	−0.08, 0.00		−0.04		0.03	−0.11, 0.03
Grade of STRS assessment	0.00	0.01	−0.02, 0.02		−0.03	0.01	−0.07, 0.02		0.01		0.02	−0.04, 0.05
Method concordance	−0.23	0.07	−0.38, −0.08		−0.18	0.08	−0.42, 0.06		−0.21		0.11	−0.48, 0.06
Conflict	−0.21	0.06	−0.33, −0.08		0.22	0.16	−0.53, 0.97		−0.03		0.08	−0.21, 0.16
Closeness	−0.17	0.06	−0.29, −0.05		0.13	0.13	−0.67, 0.93		−0.09		0.07	−0.28, 0.10
Dependency	−0.37	0.07	−0.52, −0.23		0.05	0.16	−0.38, 0.48		−0.15		0.09	−0.36, 0.05

Note. b = unstandardized beta coefficient. SE = standard error. CI = confidence interval. STRS = Student-Teacher Relationship Scale.

¹k = 237.

²k = 95.

³k = 129.

⁴Different reporters/methods = 1, same reporter/method for both measures = 0.

⁵Total STRS score is presented as the comparison group. For social skills, conflict was associated with a significantly larger effect than dependency (b = −0.16, SE = 0.05, 95% CI: −0.27, −0.05). Closeness was associated with a significantly larger effect than dependency (b = −0.20, SE = 0.05, 95% CI: −0.30, −0.09). There was no significant difference between conflict and closeness (b = 0.03, SE = 0.02, 95% CI: −0.02, 0.09). For peer relationship quality and social skills, no significant differences were detected between any STRS scales.

Measurement

A binary indicator was created to represent whether the STRS and the corresponding social competence measure for each effect were completed by the same reporter (i.e., the same teacher or the student for both = 1; a different reporter = 0). Models including methodological predictors (Table 2, Step 3; Table 3, Step 3) suggested that using different reporters across measures was associated with smaller effect sizes for social competence in general and for social skills. Additionally, we compared effect sizes across the total score and subscales of the STRS. For social competence in general (Table 2, Step 3), the total STRS score produced larger effects than did the closeness or dependency subscales but was not significantly more associated with social competence than was the conflict subscale. Furthermore, the conflict and closeness subscales were more strongly associated with social competence than was the dependency subscale. When examining effect sizes across subscales for each domain of social competence (Table 3, Step 3), a similar pattern of results emerged for social skills (i.e., total scores produced stronger effects than did the conflict, closeness, or dependency subscales, and conflict and closeness produced stronger effects than did the dependency subscale). However, there were no significant differences across subscales detected for peer relationship quality or social acceptance. This overall pattern of results did not change when all other predictor variables were included in the final model (Table 2, Step 4; Table 3, Step 4).

Discussion

Implications for Researchers

The goal of this meta-analysis was to systematically collect and review literature reporting a correlation between the STRS and social competence from preschool through high school. After extensive database searching and screening, 87 research studies were identified as relating to the topic and containing sufficient information for inclusion in the present meta-analysis. Using three-level meta-regression, overall effect sizes of r = .31 (overall social competence), r = .36 (social skills), r = .28 (peer relationship quality), and r = .28 (social acceptance) were estimated. The correlations for social competence broadly and social skills can be described as large (i.e., .30 ≤ r), whereas the correlations for peer relationship quality and social acceptance can be described as medium (i.e., .20 ≤ r < .30) using Funder and Ozer’s (2019) criteria for interpreting effect sizes in psychological research. More experimental studies are needed to inform our understanding of the causal mechanisms at play, and to provide evidence that interventions intending to increase peer social competence via teacher training are efficacious for students in a variety of grades and from a variety of sociocultural backgrounds.

The second research question addressed the extent to which developmental factors explained differences in effect sizes. Contrary to our hypothesis, the time spanned between assessments was not associated with effect size for social competence broadly, nor for any of the three subdomains. This finding suggests that a teacher-student relationship, as measured by the STRS, continues to play a role in a student’s developmental trajectory over long periods of time (i.e., may be more consistent with an enduring effects developmental model of the consequences of early experiences; see Magro et al., 2020 and Roisman & Fraley, 2013 for a discussion). That is, teacher-student relationship quality may continue to influence a student’s social behavior even after leaving a particular teacher’s classroom. This finding is most consistent with attachment perspectives on teacher-student relationship quality, which emphasize the particular importance of relationships early in life because they continue to have an impact on developmental trajectories, even after interactions have changed or ended. This hypothesis could be tested formally in longitudinal studies of teacher-student relationship quality that include repeated measures of both the STRS and social outcomes. Follow-up work in this domain is important given that the timing variable in the present sample was heavily skewed: 77.7% of effects in the full sample were cross-sectional. Given the limited longitudinal data available, this result should be interpreted with caution.

In contrast to our expectations, the grade at which the STRS was administered was also not related to effect size in any analysis. This finding may suggest that the STRS is an important predictor of social competence regardless of the age of the student. However, it may also be a mere reflection of our finding that the STRS is not typically used to assess teacher-student relationship quality beyond Grade 6. Indeed, once youth reach junior high school, the number of teachers they interact with increases substantially. Other measures of teacher-student relationship quality become more common in the secondary setting, such as those that rely on student-reports and that require students to reflect on their relationships with multiple teachers. It is not clear from the present meta-analysis whether the STRS can be used meaningfully with older students in light of both the paucity of research in this domain and the conceptual changes in the meaning of the teacher-student relationship for older students. Some researchers have suggested that the traditional domains of teacher-student relationship quality assessment (i.e., closeness, conflict, dependency, or attachment) are most valid with younger students, whereas other domains, such as emotional support, are more important for older children and adolescents (Sabol & Pianta, 2012). Future studies are needed that directly compare the predictive utility of the STRS against other measures of teacher-student relationship quality for younger versus older students across the full range of schooling. Without such studies, it is unclear whether attachment perspectives on the function of teacher-student relationships are valid for older children and adolescents.

The third research question addressed the extent to which methodological variables were associated with effect size. Consistent with our hypothesis, use of different informants for the STRS and social competence was associated with a smaller correlation between the two. This effect was most observed in the full dataset as well as the social skills meta-analysis, but not for peer relationship quality or social acceptance. The considerable contribution of reporter concordance to effect size in the present analysis is consistent with psychometric work that emphasizes the importance of partitioning out covariance that is due to shared methods or informants (Campbell & Fiske, 1959; Tehseen et al., 2017). Shared method variance can artificially inflate correlations between latent constructs. Future researchers who are estimating the correlation between teacher-student relationship quality and peer social competence should therefore expect smaller associations between the two when relying on different reporters and/or measures. Furthermore, researchers attempting to estimate the causal influence of teacher-student relationships on social development should strive to employ multi-method, multi-informant assessments in order to rule out shared method bias that may result in over-estimation of the effects of teacher-student relationship interventions. Indeed, relying on teacher reports alone of teacher-student relationship quality (as is typically the case when the STRS is used) undermines the importance of the child’s perspective in the dyadic relationship (Spilt & Koomen, 2022) and can mask within-student differences in relationship quality across teachers (e.g., Roorda & Bosman, 2022).

Finally, the present analyses compared the predictive validity of the subscales of the STRS for social competence. A comparison of the STRS subscales suggested that the conflict and closeness scores better predicted social competence, and specifically social skills, than did the dependency scale. Furthermore, the total score predicted social competence more strongly than did the closeness or dependency scales in the full sample and performed better than all three subscales for social skills. Importantly, these results suggest that the STRS dependency subscale in particular performs weakly in terms of predicting social competence. This finding is consistent with the recommended practice of using the newer version of the STRS, the Student-Teacher Relationship Scale—Short Form (Pianta, 2001b), which does not include a dependency subscale.

Methodological implications are also of note. Future researchers who wish to reduce their reliance on null hypothesis significance testing, as has been called for in psychological research for decades as a necessary step to enhance theory-testing and reduce false positive findings (e.g., Chambers, 2017; Meehl, 1967; Wright, 2003), may use the present meta-analytic estimates and standard errors to inform prior probability distributions in a Bayesian analysis framework (Etz & Vandekerckhove, 2018; Jaynes, 1986). The present results also suggest that accounting for the use of different reporters and use of specific subscales from the STRS can inform a priori estimates of effect size. Use of a priori estimates in future empirical studies can enhance our confidence in the validity of future results, and thus increase their utility for making educational and developmental policy recommendations, by substantially decreasing the risk of false positive findings inherent to the use of null hypothesis significance testing.

Implications for Practitioners

Taken together, the present meta-analytic findings suggest that teacher-student relationships as assessed with the STRS are associated both concurrently and prospectively with students’ social competence. Students who have difficulty interacting with their teachers in the elementary years may be at-risk for maladaptive social development in the present and future. Although the causal mechanisms underlying this association are not yet clear (i.e., it is unknown whether teacher-student relationships are a context in which social competencies can be improved, or whether more socially competent students are able to form better relationships with teachers, or both), educators can consider teacher-student relationships to be an important marker of children’s current and future social competence with peers. That is, teachers who are struggling to connect with a student or are experiencing significant relational conflict with a student may wish to assess the quality of that student’s overall social functioning with peers. The present findings suggest that such students are at risk for difficulties making friends and may be more vulnerable to peer exclusion and isolation. Likewise, educators who have concerns about a child’s social functioning with peers may also wish to reflect on the extent to which disconnectedness or conflict is present in the teacher-child relationship and seek to improve this aspect of the child’s social world.

Evidence-based interventions for improving teacher-student relationships have begun to emerge. Although understanding of the causal relation between the two constructs is limited, improving teacher-student relationships for students struggling to connect with others may in the short- and long-term help to improve their social competence with peers. Some intervention work already exists that aims to improve teacher-student relationships or teacher interactional styles in the hope of supporting positive child development (Sabol & Pianta, 2012). For example, Driscoll and Pianta (2010) showed that their Banking Time intervention increased teachers’ perceptions of closeness with the students in their classroom and also improved teacher-reported social competence. More recently, the Establish-Maintain-Restore approach has been employed as a method for improving teacher-student relationship quality and has resulted in improved student classroom behavior (C. Cook et al., 2018; Duong et al., 2019). More targeted interventions, such as Teacher-Student Interaction Coaching, have also been effectively used to improve relationships with children throughout the classroom by targeting teacher-student relationships that have been identified as challenging (Bosman et al., 2021).

It is also important to note that the present findings suggest that both teacher-student closeness and teacher-student conflict are associated with social competence with peers. As such, practitioners seeking to improve teacher-student relationships as a method for increasing social competence in students should focus both on increasing closeness between teachers and students and decreasing the amount of conflict in the relationship in order for students to maximally benefit from a supportive relationship with a teacher. This strategy is consistent with interventions like Teacher-Child Interaction Training, which strives first to increase positive interactions between teachers and students and later to support a teacher’s use of strategies that minimize conflict when behavioral corrections are needed (Gershenson et al., 2010; Tiano & McNeil, 2006).

Limitations and Future Directions

The present meta-analysis was limited in several important respects. First, all meta-analytic corpora are limited by the qualities of the primary studies of which they are comprised. The identified sample of studies overrepresented some groups and underrepresented others. For example, the majority of studies represented in the present analysis included preschool and elementary school students, with preschool and Kindergarten being the most heavily weighted. This skewed distribution of ages makes the precision of estimates for older students and the generalizability of these findings more limited. Future studies should explicitly test the validity of the STRS with older students. Such studies can provide valuable insight into whether conceptualizing teacher-student relationships from a developmental and attachment perspective remains relevant in the secondary grades, whether the STRS specifically remains a strong predictor of social competence over longer timespans, and the extent to which specific dimensions informed by attachment theory (i.e., closeness, conflict, and dependency) versus emotional support more broadly maintain the same meaning in older samples.

Additionally, the sample heavily overrepresented students living in the United States. Among other countries represented, the majority were European or North American. In particular, studies from South America, Africa, the Middle East, and South Asia were lacking. Importantly, the present sample was also limited by the use of English-language search terms and the limited number of manuscript languages that could be included (i.e., Chinese, English, German, and Spanish). Several studies were identified in other languages that could not be screened because translators were not available: one each in French (Lapointe & Legault, 2004), Lithuanian (Magelinskaitė, 2011), Portuguese (Coelho, 2008), and Turkish (Öz et al., 2015) were excluded after abstract screening. Given that education systems, teacher training, and expectations for social behavior in students are embedded within cultural norms that vary greatly both within and between countries (Bronfenbrenner & Morris, 2007; Marks, 2005), greater representation from outside of the United States is needed to understand the generalizability of these results. Indeed, some studies have provided evidence that student-teacher relationships have different meanings and consequences in different cultural contexts (Fredriksen & Rhodes, 2004; Joshi, 2009). Therefore, we cannot fully understand the nature of teacher-student relationships or their consequences without considering the contexts in which they are formed (Verschueren & Koomen, 2012).

The present study explicitly sought to summarize the literature on the STRS, which reflects a developmental and attachment perspective on affective qualities of teacher-student relationships. However, important work regarding teacher-student relationship quality has also been conducted from other theoretical perspectives and is reflected in other constructs such as teacher support (e.g., Hoy & Weinstein, 2006), teacher-student affiliation (e.g., den Brok et al., 2006; Fraser & Walberg, 2005), and communication and proximity between teachers and students (e.g., Wubbels & Brekelmans, 2005). The present findings may not generalize to other measures of teacher-student relationship quality. As such, future meta-analytic work that focuses on the predictive significance for social competence of other key measures, such as the Questionnaire on Teacher Interaction (den Brok et al., 2006), will help contextualize the current findings within the broader teacher-student relationship quality literature.

Another limitation of the present literature is its reliance on bivariate associations between the STRS and social competence. Although the present meta-analysis lends support to the notion that the STRS is associated with peer social competence (including social skills, peer relationship quality, and social acceptance), these associations tell us little about causality or the mechanisms that might explain them. In fact, students with higher social competence may be more likely to form positive relationships with their teachers, reflecting an alternative unidirectional or a bidirectional relation between teacher-student relationship quality and social competence. Indeed, recent evidence suggests that there is a reciprocal relation over time between teacher-student relationship quality and social skills, such that students with more self-control are likely to have lower levels of conflict in later grades, and vice-versa (Hajovsky et al., 2021). Furthermore, teacher-student closeness predicts higher interpersonal skills, but both interpersonal skills and self-control predict higher closeness (Hajovsky et al., 2021). Experimental work intending to increase the quality of teacher-student relationships, as well as experiments that seek to provide students with enhanced social skills, are needed to inform researchers and policymakers about possible causal relations between the two constructs. Studies that disaggregate within-person changes from between-person differences in teacher-student relationship quality and social competence over time can also shed light on the specific direction, or bidirectionality, of the relation (Berry & Willoughby, 2017; Curran & Bauer, 2011). Such longitudinal studies also have the potential to uncover various mediators that may help explain the ways in which teacher-student relationships and social competence with peers are influenced and influence one another (Roisman et al., 2016).

Finally, the present meta-analysis examined whether certain sample characteristics (e.g., age) moderated the relation between teacher-student relationship quality and social competence. However, further investigation within samples is needed to ascertain whether teacher-student relationships have a greater influence on social development for some students than others. For example, there is some evidence to support the notion that early adversity moderates the relation between teacher-student relationships and later developmental outcomes (e.g., achievement; Liew et al., 2010), but no work to date has specifically examined the impact on social competence. Identifying factors at the level of individuals, classrooms, and schools that are associated with teacher-student relationship quality is crucial for future intervention and professional development work (Sabol & Pianta, 2012).

Conclusion

The present study provides evidence that teacher-student relationships as measured by the STRS are both correlated with and predictive of students’ social competence with peers in the domains of social skills, peer relationship quality, and social acceptance. These associations appear to reduce when different reporters are used for the STRS and social competence. Additionally, effect sizes were larger when the total score from the STRS was used and smallest when the dependency subscale was used. More research is needed to enhance the generalizability of these findings regarding the STRS and causal inferences, including work with samples outside of the United States, use of the STRS with older students, and experimental designs and statistical innovations that enhance understanding of causal relationships.

Impact statement:

Closer and less conflictual teacher-student relationships are consistently associated with higher social skills, peer relationships, and social acceptance among peers from early childhood through adolescence. Further work implementing teacher training programs that aim to improve teacher-student relationship quality as a mechanism for enhancing students’ social competence with peers is needed to evaluate the effectiveness of such trainings across ages and sociocultural contexts.

Data availability statement:

Data for this analysis are available at https://osf.io/c97wu/