Abstract
Executive functions (EFs), important cognitive processes that enable goal-directed behavior, develop due to maturation and environmental stimulation. The current study systematically reviews and synthesizes evidence on the association between teacher–student interactions and EFs. The search resulted in 28 studies, from which 23 studies provided sufficient data to be included in the calculations. Overall effect sizes indicate that teacher–child interactions are related to general executive functioning, working memory, and inhibition but not cognitive flexibility. Relationships were stronger for studies including children at the beginning of elementary school, studies with higher socioeconomic status participants and more boys, and studies measuring teacher–child interactions at the dyadic level. This study shows that qualitative teacher–child interactions are important for performance in EFs in children. This suggests that teachers can promote the cognitive processes that are essential in children’s learning by changing their behavior to create an emotionally positive, structured, and cognitively stimulating classroom environment.
Successful functioning throughout life (e.g., in school, social interactions, physical and mental health, and professional careers) requires, among other things, well-developed executive functions (EFs). EFs refer to the cognitive processes needed to regulate behavior, thoughts, and emotions (Huizinga, Dolan, & van der Molen, 2006; Zelazo & Carlson, 2012). The importance of EFs for a large variety of life domains has driven an increase in studies exploring EFs development and how this development can be improved. Recently, positive interactions between teachers and their students have been related to improvements in development of EFs (e.g., Berry, 2012; de Wilde, Koot, & van Lier, 2016; Hamre, Hatfield, Pianta, & Jamil, 2014). However, research on this topic is fragmented due to the inconsistent use of terminology, definitions, and operationalization of both the construct of EFs and teacher–student interactions. The current study systematically reviews existing literature and synthesizes results into a meta-analysis to provide an overview of the available research.
Executive Functioning
EFs refer to a complex construct that has been defined in many different ways. Nevertheless, these definitions have some common elements emphasizing the key characteristics of EFs. First, EFs are defined as cognitive processes exerting top-down influence over actions, thoughts, and emotions (Zelazo & Carlson, 2012). Second, EFs are involved in conscious, goal-directed behavior and not in situations that rely on automatized or intuitive behavior (Huizinga et al., 2006; Zelazo & Carlson, 2012). Consequently, using EFs requires effort. Third, EFs refer to a set of interconnected, though distinguishable, cognitive processes and is thus considered to be a multidimensional rather than a unitary construct (Diamond, 2013; Garon, Bryson, & Smith, 2008; Zelazo & Carlson, 2012). This multidimensionality of EFs is also confirmed for young children in numerous empirical studies (Huizinga et al., 2006; Letho, Juujärvi, Kooistra, & Pulkkinen, 2003; Miller, Giesbrecht, Müller, McInerney, & Kerns, 2012; Usai, Viterbori, Traverso, & De Franchis, 2014). Most researchers agree that there are three core EFs (i.e., working memory, inhibition, and cognitive flexibility) that develop during the first years of life (Diamond, 2013; Huizinga et al., 2006). At a later stage these core EFs are integrated to support more complex EFs, such as planning and reasoning (Diamond, 2013). In the current study, we focus on the three core EFs in a classroom context.
Working memory refers to the memory system that consist of two components to temporarily store information (the phonological loop and visuospatial sketch pad) and one component for the processing or manipulating information (the central executive; Baddeley, 1992). For example, doing mental mathematics requires to remember the numbers and operators and to make the calculations. Within the classroom context, working memory is essential in children’s learning in preschool and elementary school as shown by positive relationships between working memory and children’s school engagement (Fitzpatrick & Pagani, 2012), following instructions in the classroom (Gathercole, Durling, Evans, Jeffcock, & Stone, 2007), and academic achievement (Alloway & Alloway, 2010; Friso-van den Bos, van der Ven, Kroesbergen, & van Luit, 2013). Additionally, working memory is important for better social functioning (McQuade, Murray-Close, Shoulberg, & Hoza, 2013) and adequate emotional functioning (Brunnekreef et al., 2007; Engen & Kanske, 2013; Joormann & Gotlib, 2008).
Inhibition is the ability to suppress an automatic response, thought, or emotion in favor of a less dominant, though often more appropriate, response (Bari & Robbins, 2013). For example, when working on an important task attention needs to be focused on the relevant aspects of the task, while filtering out distracting sounds or thoughts. In an extensive longitudinal study, Moffitt et al. (2011) linked the ability to resist temptation at the age of 3 to positive outcomes in several domains, such as physical health and socioeconomic status (SES) in adulthood. Within the elementary classroom context, inhibition has been shown to be important for behaviors important for learning, such as attention and externalizing behaviors (Berry, 2012), social skills and internalizing behavior (Rhoades, Greenberg, & Domitrovich, 2009), and learning outcomes, such as academic achievement (Blair & Razza, 2007; Lubin, Regrin, Boulc’h, Pacton, & Lanoe, 2016).
Cognitive flexibility refers to viewing matters from diverse perspectives and efficiently handling these perspectives in order to adapt to the changing demands of the situation (Canas, Quesada, Antoli, & Fajardo, 2003). All situations in which (unexpected) changes occur require cognitive flexibility. For example, when students need to change from one activity to another in the classroom, they need to turn their attention away from the previous activity, let go of their old response set, and shift their attention and response set to meet the demands of the new activity. Children with higher performance on measures of cognitive flexibility generally have better social understanding (Bock, Gallaway, & Hund, 2015) and better reading skills (Cartwright et al., 2017) in middle childhood.
As the three core EFs each play a crucial role in children’s development, classroom functioning, and learning, it is important to understand how these functions develop and which malleable factors in children’s proximal environments can promote this development.
EFs Development, Plasticity, and the Role of the Environment
EFs are related to different brain regions, the most important being the prefrontal cortex (PFC; V. Anderson, Jacobs, & Anderson, 2008; McKenna, Rushe, & Woodcock, 2017). EFs development parallels changes in this brain region and is characterized by alternating periods of rapid and gradual growth (V. Anderson et al., 2008). Elementary forms of behavioral control emerge in infants (before the age of 2), when neural density in the PFC increases (Diamond, 2002). Between the age of 2 and 7 the neural pathways in the PFC are improved and function more efficiently due to synaptic pruning processes (Diamond, 2002; Zhong et al., 2014). During this period an important developmental spurt occurs in which the three core EFs develop rapidly (Zelazo & Carlson, 2012). Both the PFC and the three core EFs continue to progress at a more gradual rate until adolescence or even young adulthood (P. Anderson, 2002; Diamond, 2002; Zelazo & Carlson, 2012).
Although maturation is important for the development of EF, the brain shows remarkable anatomical and functional plasticity, especially in periods of rapid growth (V. Anderson et al., 2008; McEwen & Morrison, 2013; Zelazo & Carlson, 2012). Due to the prolonged development of the prefrontal brain regions, with alternating periods of rapid and gradual growth, there is a long window for environmental stimulation to promote EFs development (Hughes, 2011). Studies on the possibility to train EFs have shown that effects of computerized training do not generalize to real-life classroom situations, suggesting that performance in EFs can be improved in young children (e.g., Thorell, Lindqvist, Nutley, Bohlin, & Klingberg, 2009), though focusing only on EFs does not improve related behavior and outcomes in the classroom context (Diamond & Lee, 2011; Melby-Lervag & Hulme, 2013). Including children’s developmental contexts in interventions targeting EFs is needed. Programs that include stimulation of EFs and also pay attention to the classroom contexts have indeed shown better transfer effects (e.g., Bierman, Nix, Greenberg, Blair, & Domitrovich, 2008; Raver et al., 2011).
It can be assumed that in the course of normal EFs development, outside of intervention situations, stimulation provided in children’s environments play a role as well. Compared to the literature on the biological underpinnings of EFs and the effects of training, the research examining the role of the environment for performance in EFs and development is more recent, and the studies are limited in number (Hughes, 2011). Most available studies focus on the family and home environment as an important developmental context for EFs, showing that positive parent–child interactions can stimulate EFs (see Hughes, 2011, for an overview). However, when children enter formal schooling, the classroom context and interactions with teachers also become an important part of their environment and developmental context (Bronfenbrenner & Ceci, 1994; Downer, Sabol, & Hamre, 2010). So far, the role of teacher–child interactions in EFs performance has rarely been investigated.
Teacher–Child Interactions and EFs
There are different ways of looking at the teacher–student interaction. In general teacher–child interactions are studied at two levels, the dyadic level and the classroom level. In young children, an attachment-based perspective is often used in research (Sabol & Pianta, 2012; Verschueren & Koomen, 2012). This framework focuses on the affective components of the dyadic relationship between the teacher and a student. More specifically, it looks at closeness, conflict, and dependency in the relation between the teacher and a specific student (Koomen, Verschueren, van Schooten, Jak, & Pianta, 2012; Verschueren & Koomen, 2012). Closeness refers to the degree of affection, warmth, and open communication between the child and teacher, whereas conflict includes the degree of negative affect expressed and experienced in the relationship (Ahnert, Harwardt-Heinecke, Kappler, Eckstein-Madry, & Milatz, 2012; Hamre & Pianta, 2001). Dependency refers to clingy behavior of the child toward the teacher and the overreliance on the teacher (e.g., continuously asking for help, even when it is not needed; Ahnert et al., 2012; Hamre & Pianta, 2001; Koomen et al., 2012; Vervoort, Doumen, & Verschueren, 2015). However, such a dyadic relationship does not stand on its own. It is embedded within the broader classroom context, which can influence children’s development in its own way (Howes et al., 2011). It is therefore important to also consider the interactions of the teacher with the class as a group. At the classroom level, research has mainly focused on teacher’s emotional support, classroom organization, and instructional support as important factors for children’s classroom functioning (Downer et al., 2010; Hamre et al., 2014). Emotional support refers to behaviors that promote the affective quality of interactions with the teachers and that increase children’s comfort and sense of security within the classroom (e.g., smiling, acknowledging children’s emotions and experiences and sensitively responding to it, allowing children to speak their mind). Behaviors that contribute to the smooth functioning of the classroom are classified in the domain of classroom organization (e.g., clarifying rules and expectations, minimizing transition time, proactive behavior management). Finally, instructional support includes those interactions that promote children’s cognitive and higher order thinking, such as asking open-ended questions, providing feedback that extends children’s learning, and providing opportunities to use advanced language.
Although previous studies have clearly shown that the classroom environment, and in particular the interactions between students and teachers, can influence children’s development (Downer et al., 2010; Verschueren & Koomen, 2012), research examining the role of this environmental context for children’s EFs has only recently gained attention. The first studies are now showing that positive relationships and interactions with teachers can promote EFs, whereas negative relationships can hinder EFs (e.g., Berry, 2012; de Wilde et al., 2016). These results are promising, yet evidence is hard to interpret for three reasons. First, research examining the relationship between teacher–child interactions and EFs has operationalized teacher–child interactions in different ways. Some studies have examined these interactions at the dyadic level (e.g., de Wilde et al., 2016), while others have investigated these interactions at the classroom level (e.g., Rimm-Kaufman, Curby, Grimm, Nathanson, & Brock, 2009). It is unclear what the relative importance of each of these levels is for children’s EFs. Additionally, there are different operationalizations of these interactions. For example, teacher–child interactions are measured from both the perspective of the teacher (e.g., Cadima, Doumen, Verschueren, & Buyse, 2015) and the perspective of the child (e.g., de Wilde et al., 2016). Second, different researchers define and measure EFs in a variety of ways and some concepts that are related to but different from EFs are used interchangeably (e.g., self-regulation, effortful control; see Nigg, 2016, for an overview of terminology). Third, there are differences in the study designs examining the relation between teacher–child interactions and EFs. In particular, one type of studies uses cross-sectional designs and the other type aims to predict performance in EFs based on teacher–child interactions measured at an earlier point in time. Altogether, these differences make the research on this topic fragmented and make it hard to draw conclusions across studies. The current study tries to fill this gap.
Research Questions
The current study examines the relationship between teacher–student interactions and EFs by means of a systematic review and meta-analysis. Insights into the association between these concepts is essential for understanding how teacher–student interactions might be used to prevent or decrease problems with EFs and related academic issues. If we can understand which aspects of these interactions are important for which EFs, we can attempt to improve specific teacher behaviors in order to target problems with specific EFs. For example, if the amount of warmth in the relationship between teachers and students is the most important, then focusing on increasing teachers’ sensitivity is likely to be more effective than improving classroom organization when it comes to improving EFs. Focusing on the quality of teacher–student interactions could then be complementary to training of EFs and might increase transfer effects of such trainings.
The current study has three goals:
To provide an overview of the existing research on the association between teacher–student interactions and children’s EFs;
To provide an estimate of the mean effect size of the association between teacher–child interactions and EFs;
To examine whether the effect size depends on sample characteristics (age, gender, SES), the level of teacher–student interactions (dyadic vs. classroom level), the dimensions of the teacher–student interactions (e.g., closeness, conflict, classroom organization) measured, and the study design (cross-sectional vs. longitudinal).
These questions were addressed with a systematic search and a meta-analysis in which we summarize the empirical evidence on the relationship between different aspects of teacher–student interactions and EFs.
Method
Search
First, a database search was conducted in Web of Science, PubMed, and ERIC. Because both teacher–student interactions and EFs are complex constructs comprising multiple dimensions and components, an extensive list of search terms was used (Table 1), based on the literature (Diamond, 2013; Nigg, 2016). A first selection of studies was made based on a comparison of the title and abstract with the inclusion and exclusion criteria (see below). If an exclusion criterion was applicable, then the study was excluded from further review. If all inclusion criteria were applicable or the title and abstract contained insufficient information for exclusion or inclusion, then the study was included. From there, the included studies were compared to the inclusion and exclusion criteria again, this time based on the full text.
Search terms used for the database search in Web of Science, Pubmed, and Eric
Second, those publications that were included based on the database search were used for a backward and forward search. For the backward search, the reference lists of all included documents were inspected. The titles (and abstracts if necessary) were compared to the eligibility criteria. New papers that met inclusion criteria were added to the sample of studies. For the forward search, all papers were examined that referred to one of the papers already included. Once these papers were identified, the papers were again compared to the inclusion criteria and included in the sample if appropriate.
The full process of the search for and selection of studies was conducted by two people, the first author and the fourth author. After the selection based on title and abstract, results from the two researchers were compared. There was an original agreement of 97.46% in the screening phase and of 90% for the reading of the full text. Studies for which there was disagreement were discussed until an agreement was reached.
Inclusion and Exclusion Criteria
Only studies reported in English were considered for the current review. Studies were included if the following criteria were met: (1) the sample consisted of children between the ages of 2 and 12, (2) the study comprised a community sample, (3) teacher–child interactions were measured, and (4) EFs were measured.
The age of the participants included in the study ranged from 2 to 12 years. This age range corresponds to the period in which children start preschool (Bertram & Pascal, 2002) until the end of primary education (Le Métais, 2003) in most countries. Children from secondary education and older were excluded as they are often exposed to multiple teachers throughout the day (more so than in primary education) and therefore teacher–child interactions may have a different meaning and a different effect on student outcomes at this older age (Roorda, Koomen, Spilt, & Oort, 2011). Similarly, we focused on children recruited in community samples, and not clinical samples, as the studied processes may differ for children with a medical or psychiatric diagnosis. For example, studies that focus on samples of children with attention deficit/hyperactivity disorder or autism spectrum disorder were excluded. The study was also included if data were available on a subsample that corresponded to the inclusion criteria.
The study had to include a general measure for EFs or specifically assess one of the three core EFs, working memory, inhibition, and cognitive flexibility. These measures of EFs need to be clearly distinguished from measures capturing related concepts, such as self-regulation or effortful control. The term self-regulation is sometimes used interchangeably with EFs (e.g., Morrison, Ponitz, & McClelland, 2010), seen as a part of EFs (e.g., Diamond, 2013), or seen as a broader construct that includes EFs (e.g., Nigg, 2016). Similarly, effortful control is a concept that is also sometimes used as an equivalent of EFs (e.g., Eisenberg, Smith, & Spinrad, 2011), whereas others assume it is an aspect of temperament, has a more constitutional basis, and is more stable than EFs (e.g., Neuenschwander, Rothlisberger, Cimeli, & Roebers, 2012). Because the use of these terms often differs between studies and researchers, the decision to include or exclude a study was based on the specific measurement used. When tasks were used that explicitly involved the use of top-down controlled cognitive processes or behavior the study was included. Indirect measures used, such as questionnaires on temperament, impulsivity, self-regulation, or EFs (e.g., Barbarin, 2013; Cappella et al., 2012), led to exclusion. Additionally, a measure for teacher–student interactions, at either the dyadic or classroom level, had to be included. Measures assessing the teacher–child interactions had to indicate the quality of those interactions and not merely the amount of time spent in interactions (e.g., the amount of time spent in individual instruction). Studies were excluded when the measure of teacher–student interactions and/or EFs was confounded with other measures (e.g., a general cognitive measure combining EFs with other cognitive processes). Finally, the study was included only if it provided data where teacher–child interactions were assessed before or simultaneously with the EFs measure, as this approach allows us to answer the current research questions.
Sample
Figure 1 gives an overview of the number of studies that were found, included, and excluded in each step of the process. The database search resulted in a sample of 2,167 studies. The database and backward and forward search resulted in the inclusion of 28 original empirical studies. Additionally, we found one review study and one book chapter including a section on teacher–child interactions and EFs (Blair & Raver, 2015; McGrath, Thurman, Raisch, & Lucey, 2016) and one meta-analysis on the associations between observations of teacher–child interactions at the classroom level and child outcomes, including some EFs outcomes (Perlman et al., 2016). These were used for a backward and forward search. For these 28 studies statistical data were extracted and or requested from the author(s). For 23 studies sufficient statistical information was collected to calculate an effect size. Most studies were reported in articles published in journals, three studies were found in the form of conference abstracts, and two studies were part of a PhD dissertation.
Flowchart of study search and inclusion. Flowchart indicates the number of studies found in the database search, studies found in the backward search and forward search, and studies excluded and included during the screening of titles and abstracts and the reading of the full text.
Data Collection and Preparation
Data were extracted from the studies. Sample characteristics, study characteristics, and statistics were noted and recoded for use in the analyses. With regard to the sample characteristics the number, mean age (in months), and gender (percentage of boys) of participants was documented. Additionally, information on the SES was gathered for the participating children (parental education, family income, parental work status, and/or eligibility for reduced lunch fees) and/or the setting in which the study was conducted (educational level, average income, unemployment rates, or general rates on reduced–lunch fee eligibility for the school or geographical area). If a high number of participants (>50%) had a low SES or the studied area was characterized by low SES (compared to the average of the country as mentioned in the included paper), then the study was coded as examining a low-SES sample.
With regard to the study characteristics, data were collected on the study design, the dimension of teacher–child interactions that was measured, and the core EF that was measured. With regard to the study design, the effect sizes were categorized as either cross-sectional or longitudinal, depending on the timing of the measures. The aspect of the teacher–child interactions that was examined was coded as closeness, conflict, dependency, general dyadic relationship, emotional support, organizational support, instructional support, and general classroom interactions. These concepts were chosen as they are supported by the most commonly used models for characterizing teacher–child interactions (see above). The type of measure (questionnaire or observation) and the name of the measure were also documented. A similar approach was used for the measurement of EFs. Based on the definitions of working memory, inhibition, and cognitive flexibility in the introduction, it was determined in which category a measure was included. When the measure focused on remembering and/or processing information over a short period of time, it was classified as a working memory measure. Measures of long-term memory or episodic memory were not included. If the main focus of the measure was on the degree to which the child could suppress a dominant thought, emotion, or response, it was labeled inhibition. Measures were categorized as cognitive flexibility assessments when the child had to generate ideas flexibly or shift between different cognitive or behavioral sets. When a measure included more than one of these processes or different measures were combined in a composite score, then it was categorized as general EF. Again, the type of measurement (test, questionnaire, observation) and the name of the instrument were noted.
Finally, the statistical information needed to calculate the association between the teacher–child interactions and EFs was recorded and transformed into a correlation. In case the information needed to compute the correlation was missing, the authors were contacted by email and the information was requested. Correlations were always registered in such a way that positive correlations indicated that better teacher–child interactions were related to better EF performance.
Data Analyses
Analyses
First, a qualitative analysis of the systematic review is reported, providing an overview of the existing literature, characteristics of the included studies, and their conclusions.
Next, meta-analysis is used to calculate the mean effect size of the association between teacher–child interactions and EFs. Since this study did not examine differences between conditions or groups but rather studied the association between two (continues) variables, unconditional Pearson correlations are used as an effect size. From the 28 studies, 14 studies reported the correlations (Abenavoli & Greenberg, 2014; Berry, 2012; Cadima et al., 2015; Cadima, Verschueren, Leal, & Guedes, 2016; Conradt et al., 2014; de Wilde et al., 2016; Finch, Johnson, & Phillips, 2015; Fuhs, Farran, & Nesbitt, 2013; Hong, 2012; Liew, Chen, & Hughes, 2010; Rimm-Kaufman et al., 2009; Spilt & Hughes, 2015; Valiente, Swanson, & Lemery-Chalfant, 2012; Williford, Maier, Downer, Pianta, & Howes, 2013). For one study the correlations were calculated based on means and standard deviations that were provided for children scoring high and low on secure-base behavior (Commodari, 2013). The 13 remaining studies reported only the results of multilevel linear analyses, and correlations were requested from the authors. Eight authors provided us with the correlations (Araujo, Carneiro, Cruz-Aguayo, & Schady, 2016; Bailey, Denham, Curby, & Bassett, 2016; Cadima, Enrico, et al., 2016; Choi et al., 2016; Hamre et al., 2014; Leyva et al., 2015; Ursache, Blair, Bierman, & Nix, 2011; Weiland, Ulvestad, Sachs, & Yoshikawa, 2013). The remaining studies were not included in the meta-analysis (Ertürk Kara, Gönen, & Pianta, 2017; Jones & Bailey, 2014; Jones, Bub, & Raver, 2013; Slot, Mulder, Verhagen, & Leseman, 2014; Yildiz, Ertürk Kara, Tanribuyurdu, & Gönen). Correlations are transformed into Fisher z coefficients for use in all calculations and are transformed back to aid interpretation of results (Borenstein, Hedges, Higgins, & Rothstein, 2009; Lipsey & Wilson, 2001). For the calculation of effect sizes, fixed or random models can be used. Random effects models are used when effect sizes of studies with different characteristics are synthesized. They provide the opportunity to generalize results (Borenstein et al., 2009). As such, these models are the most appropriate for the current data. However, random effects models have lower power to detect significant effects compared to fixed effects models (Borenstein et al., 2009). Therefore, both models will be calculated and reported in the current study. Fixed effects models are calculated when at least two studies are available (Borenstein et al., 2009). Random effects models are calculated when at least five studies are available, as these models require sufficient studies to calculate between study variation (Borenstein et al., 2009). Effect sizes are calculated across all EF outcomes but also specifically for the different outcomes (general EF measure, working memory, inhibition, and cognitive flexibility).
Finally, a meta-regression is performed to examine whether effect sizes differ for samples with different characteristics (age, gender, SES), dyadic versus classroom-level teacher–child interactions, the aspect of teacher–child interactions measured (e.g., closeness vs. conflict), and cross-sectional and longitudinal associations. To this end, the effect sizes of all types of EF outcomes are used, as most core EFs have not been studied by a sufficient number of studies to examine these separately in a meta-regression.
Dependency of Effect Sizes
In the current study we need to take into account the complex data structure that arises from the fact that most studies examined multiple aspects of teacher–student interactions in relation to EFs or measured EFs at multiple time points. As a consequence, most studies contribute more than one effects size to the meta-analysis and meta-regression. The effect sizes of these studies are thus not independent from one another, and if this issue is not addressed in the analysis, it will lead to underestimation of the variance of the mean effects, which can bias interpretation (Hedges, Tipton, & Johnson, 2010). In the current study, the weights were corrected when calculating the mean effect sizes by multiplying the inverse of the variance of the effect size with the inverse of the number of effect sizes provided by the study (Hedges et al., 2010). As a result, samples with more participants will have more weight, though each study will contribute only once in the calculation of the mean effect sizes. For the meta-regression, robust variance estimation is used to correct for the dependency between the effect sizes (Hedges et al., 2010). This approach requires 20 to 40 studies and on average more than one effect size per study (Hedges et al., 2010). The current study includes 23 studies with an average of more than five effect sizes per study.
Publication Bias
Publication bias, resulting from a tendency to publish significant results more easily than null findings, is a clear issue in educational and psychological research (Polanin, Tanner-Smith, & Hennessy, 2016). Publication bias can be (partially avoided) by also including nonpublished work in a meta-analysis (Polanin et al., 2016). Therefore, the current study also includes unpublished conference abstracts and dissertations in the analysis. However, it is always possible that other gray literature exists on the topic that was not found by the researchers.
To statistically assess publication bias, the correlation was examined between sample size and effect size. If publication bias is present, a negative correlation between the sample size and effect size should occur, because studies with smaller samples that indicate small or nonsignificant results are more likely not to be published than studies with larger samples with nonsignificant results (Kühberger, Fritz, & Scherndl, 2014; Roorda et al., 2011). In our sample we found a nonsignificant correlation (r = .02, p = .873), indicating there was no such publication bias in this study.
Results
Sample and Study Characteristics
An overview of study and sample characteristics included in the review and meta-analysis can be found in Table 2. The oldest study included was published in 2009, but most studies were published from 2013 onward. With regard to the qualitative review, the sample of 22,360 children (range = 80–7,978) had an average mean age of 61.54 months across the studies. Boys and girls are approximately equally represented in the sample (51% boys). Not all studies reported the number of classrooms, but those studies that did included a total of 2,263 classrooms. Eighteen studies had a sample with high numbers of children with a low socioeconomic background (based on parental education, family income, or eligibility for reduced lunch fees). Studies were conducted in Belgium, Chile, Ecuador, Italy, the Netherlands, Portugal, Turkey, and different parts of the United States of America.
Overview of study characteristics
Note. SES = socioeconomic status. Studies preceded by superscript a were also included in the meta-analysis. For the other studies correlations were not available, and these were thus only considered for the qualitative review. For SES, 1 = sample is characterized by high numbers of children with a low socioeconomic background)
Of these 28 studies, 23 had sufficient statistical information to calculate an effect size that could be included in the meta-analysis. In total, 19,906 children from a total of at least 1,929 classrooms were included. Mean age of this subsample was 62.91 months and 50.7% were boys. Fifteen samples were characterized as low SES; for one study, data on SES were missing.
Systematic Review
Overall EFs
Of the 28 studies included, 5 studies (17%) reported the association of some aspect of teacher–child interactions with a general measure of EF. One study used the Head-Toes-Knees-Shoulders (Cadima, Verschueren, et al., 2016) as a general EF measure. The other four studies used a composite score comprising different components of EFs (Araujo et al., 2016; Cadima, Verschueren, et al., 2016; Conradt et al., 2014; Fuhs et al., 2013). One study examined the role of dyadic teacher–child interactions, three studies investigated the teacher–child interactions at a classroom level, and one studied investigated both levels.
With regard to dyadic teacher–child relationships, a general positive dyadic relationship was significantly related to better EFs measured simultaneously (Conradt et al., 2014). At this dyadic level, results of the study of Cadima, Verschueren, et al. (2016) suggested that closeness was positively correlated to general EFs measured later, whereas conflict showed no significant relationship. With regard to the classroom level, teacher–child interactions were positively related to general EFs, even after controlling for additional variables, such as, age, gender, SES, and teacher experience (Araujo et al., 2016; Cadima, Verschueren, et al., 2016). When looking at emotional support, classroom organization and instructional support as separate aspects of classroom teacher–child interactions, there was no association with general EFs (Cadima, Verschueren, et al., 2016). Fuhs et al. (2013) showed that teachers’ behavior approval was positively related to EFs, whereas other aspects of emotional support were unrelated, and results of Cadima, Verschueren, et al. (2016) indicate that instructional support has a positive effect on EFs only for children who experience difficulties in EFs.
Working Memory
Seven studies (24%) reported on the association between teacher–child interactions and children’s working memory. A number of different measures were used to asses working memory, mainly verbal in nature: word span backward (Abenavoli & Greenberg, 2014; Ursache et al., 2011), digit span forward (Commodari, 2013; Weiland et al., 2013), digit span backward (Commodari, 2013; Hamre et al., 2014; Weiland et al., 2013; Williford et al., 2013), and dot matrix backward (de Wilde et al., 2016). Three studies examined dyadic teacher–child relationships and four studies examined teacher–child interactions at the classroom level.
With regard to the dyadic teacher–child interactions, results are mixed. Commodari (2013) revealed no association between the general affective teacher–child relationship and working memory. Results of Abenavoli and Greenberg (2014) suggest that higher levels of teacher reported closeness were related to better working memory and teacher-reported conflict is not related, whereas de Wilde et al. (2016) found a negative effect of child-reported conflict on working memory and a positive association between child-reported closeness and working memory. At the classroom level, a general measure of teacher–child interactions was related to working memory (Williford et al., 2013). Weiland et al. (2013) found no correlation with the specific dimensions of emotional support, classroom organization, and instructional support, whereas Hamre et al. (2014) found only classroom organization to be positively related with working memory but not emotional or instructional support.
Inhibition
Most studies that were found include a measure of inhibition (23 studies, 79%). Most studies used a measure tapping behavioral forms of inhibition: Preschool Self-Regulation Assessment (Bailey et al., 2016; Cadima, Enrico, et al., 2016; Ertürk Kara et al., 2017; Hong, 2011; Jones & Bailey, 2014; Jones et al., 2013; Rimm-Kaufman et al., 2009; Yildiz et al., 2014), Peg Tapping (Abenavoli & Greenberg, 2014; Choi et al., 2016; Hamre et al., 2014; Hatfield, Burchinal, Pianta, & Sideris, 2016; Leyva et al., 2015; Ursache et al., 2011; Weiland et al., 2013; Williford et al., 2013), Continuous Performance Test (Berry, 2012; Valiente et al., 2012), Circle and Star (Cadima et al., 2015), Go/No-Go task (Finch et al., 2015), and Walk-a-Line Slowly (Liew et al., 2015; Spilt & Hughes, 2015; Ursache et al., 2011). There were only two studies that used a more cognitive measure of inhibition, namely, a visual and/or auditory search task requiring selective attention (Commodari, 2013; Slot et al., 2014).
Dyadic teacher–child interactions were assessed in seven studies. Sixteen studies reported associations with classroom level interactions. With regard to the dyadic level, general measures of teacher–child relationship quality were positively associated with inhibition outcomes (Commodari, 2013), also after controlling for children’s SES (Valiente et al., 2012). Conflict was negatively correlated with inhibition (Abenavoli & Greenberg, 2014; Berry, 2012; Cadima et al., 2015), except in the study of Spilt and Hughes (2015) with an ethnically diverse sample of children where no relationship was found. Closeness was related to inhibition in only one study (Abenavoli & Greenberg, 2014), whereas two studies showed no relationship (Cadima et al., 2015; Liew et al., 2010).
At the classroom level, results are mixed. Most studies indicate no relationship between general teacher–child interactions or the specific dimensions and inhibition. A few studies do show positive associations with general teacher–child interactions (Williford, 2013), emotional support (Choi et al., 2016; Hatfield et al., 2016), classroom organization (Hamre et al., 2014; Hatfield et al., 2016), and instructional support (Hatfield et al., 2016; Leyva et al., 2015; Slot et al., 2014). One study indicated a negative association between emotional support and inhibition (Hamre et al., 2014). Finally, some studies suggest that the association might be more complex, depending on initial inhibition scores (Choi et al., 2016) or the task measuring inhibition (Yildiz et al., 2014). Weiland et al. (2013) found quadratic relationships but no linear relationships.
Cognitive Flexibility
Only three studies (11%) examined cognitive flexibility in relation to the quality of teacher–child interactions. All three studies used the Dimensional Card Change Sort task, which measures the ability to shift attention from one set of characteristics (shape) to another set of characteristics (color).
One study suggests that higher levels of closeness and lower levels of conflict are associated with better performance on the shifting task (Abenavoli & Greenberg, 2014). The other two studies examined the association with classroom-level teacher–child interactions, showing no significant associations with emotional support, classroom organization, or instructional support (Leyva et al., 2015; Ursache et al., 2011). The study of Ursache et al. (2011) only showed a positive relationship with instructional support in kindergarten, not in prekindergarten.
Meta-Analysis and Meta-Regression
An overview of the studies and effect sizes included in the meta-analysis is shown in Table 3. Table 4 shows the overview of the general effect sizes according to the fixed and random effects models. Mean effect sizes were calculated between teacher–child interactions, on one hand, and all EF outcomes, general EF, working memory, inhibition, and cognitive flexibility, on the other. The mean effect sizes were all significant, except for the effect size estimating the association between teacher–child interactions and cognitive flexibility. Effect sizes are considered as small below .10, small to medium between .10 and .25, medium around .25, medium to large between .25 and .40, and large above .40 (Lipsey & Wilson, 2001, p. 147). For all EF outcomes taken together, general EFs and inhibition the mean effect size fell in the small range. For working memory, a small to medium effect size was found.
Overview of effect sizes for the studies included in the meta-analysis
Note. TSI = teacher–child interactions; EFs = executive functions; STRS = Student–Teacher Relationship Scale; CLASS = Classroom Assessment Scoring System; TCRS = Teacher–Child Rating Scale; ORCE = Observational Ratings of the Caregiving Environment; TOP = Teacher Observation in Preschool; TSRI = Teacher–Student Relationship Inventory; YCATS = Young Children’s Appraisals of Teacher Support; NRI = Network Relationship Inventory; DCCS = Dimensional Card Change Sort; PSRA = Preschool Self-Regulation Assessment; CPT = Continuous Performance Test; HTKS = Head-Toes-Knees-Shoulders.
Overall effect sizes for the associations between teacher–student interactions and the core executive functions
Note. EFs = executive functioning, k = number of studies in the calculation, N = number of participants in the calculation. The random effects model was not calculated for the general measures of executive functions and for cognitive flexibility, due to the small number of studies.
p < .05. **p < .01. ***p < .001.
Meta-regression was conducted with effect sizes for all EF measures as the outcome and sample characteristics (age, gender, SES), the level of the teacher–child interactions (dyadic level as baseline group), and the study design (cross-sectional as baseline group) as predictors. These predictors were analyzed separately as there were not enough studies available to include all these variables in the model simultaneously. Although most pairs of moderators were unrelated, some moderators were found to be associated with one another. Specifically, in samples with higher percentages of boys mean age was higher (r = .67, p < .001), when the effect of the dyadic teacher–child relationship on EFs was investigated mean age was higher, t(112) = 7.41, p < .001, cross-sectional designs were more often used in low SES samples (χ2 = 6.12, p = .013), and the classroom level was more often examined in low-SES samples (χ2 = 13.02, p < .001). Because there where both between and within study variation for the different moderators, we examined moderator effects both across and within studies. The aspect of teacher–child interactions (e.g., closeness, conflict) could not be included because there were not enough studies available.
Results are shown in Table 5. Age and gender significantly predicted the effect size, both across and within studies. If the mean age of the sample was higher or the sample included a higher percentage of boys, higher effect sizes were found in the association between teacher–child interactions and EFs. For SES, a significant between study effect indicated that studies involving a low-SES sample reported lower effect sizes compared to the other studies. Only one study examined both a low- and a high-SES group (Cadima, Verschueren, et al., 2016). In this study, a significant relationship was found between teacher–child interactions and EFs for the low-SES group but not the high-SES group. With regard to the level of teacher–child interactions, there were significantly higher effect sizes for studies measuring dyadic teacher–child interactions when examining this moderator across studies. Within study a similar trend was visible, though this result was not significant. Finally, entering study design (cross-sectional vs. longitudinal) as a predictor resulted in a nonsignificant difference in effect sizes, both across and within studies.
Results of the meta-regression examining the moderating role of age, gender, socioeconomic status, level of teacher–child interactions, and study design in the relationship between teacher–child interactions and executive functioning
Note. All variables were entered in the regression separately, as there were not enough studies to examine the moderators simultaneously.
Discussion
The current study aimed to provide an overview on available research examining the association between teacher–child interactions and children’s EFs. A systematic review was conducted, combined with a meta-analysis and a meta-regression, to provide an estimation of the general effect size and to investigate which factors moderated this effect size.
The Association Between Teacher–Child Interactions and EFs
The mean effect sizes calculated for overall EFs and its different aspects indicated that in general positive teacher–child interactions were related to better EF performance. Only with regard to cognitive flexibility was such an association not found. However, only three studies examined this EF, and they all used the same attention-shifting task, not taking into account other aspects of cognitive flexibility, such as response-shifting or fluency. In general, the effect sizes were low to low-medium and thus indicate a modest contribution of teacher–child interactions in EF performance. Nevertheless, it should be noted that categorizing effect sizes aids interpretation of results but should be used with caution. Hill, Bloom, Black and Lipsey (2008) argue that such categorization does not take into account the context (e.g., what type of outcome measure used). The association of teacher–child relationships with other types of variables also shows small to medium effects (e.g., Roorda et al., 2011). Similarly, the associations between EFs and other variables is often of the medium range (e.g., Jacob & Parkinson, 2015). The findings of small to small-medium associations in the current study are thus in line with other research and should also be interpreted in this way.
Additionally, a meta-regression was conducted to examine whether the effect sizes observed depended on sample characteristics (average child age, child gender, and low-SES sample), research design (cross-sectional or longitudinal), the level of teacher–child interactions measured (dyadic or classroom level) or components of teacher–child interactions and EFs measured. Results indicate that the associations between teacher–child interactions and EFs were stronger for samples with older children and higher number of boys. When teacher–child interactions were measured at the dyadic level, effect sizes were stronger, though this result was significant between studies and a trend within studies. Results were more mixed for SES as a moderator, showing lower effects sizes in low-SES samples when examining between studies, while within one study a higher effect size was found in the low-SES sample compared to the high-SES sample. The research design did not affect the effect size.
With regard to age, it should be noted that all studies, except for one, examined the associations between teacher–child interactions and EFs in children between the ages of 2 and 7. They thus all examined the period in which EFs show tremendous growth (Garon et al., 2008; Zelazo & Carlson, 2012) and are likely sensitive to environmental stimulation (Diamond, 2002; Huttenlocher, 2002). It is possible that within this age-group older children are more affected by teacher–child interactions, because of the growing importance of the classroom context when children enter formal schooling. Diamond and Lee (2011) note that EF trainings are effective in 8- to 12-year-olds and classroom curricula designed to improve EFs are effective for 4- and 5-year-olds. This suggests that different factors might have their most profound impact on EFs at different points in development.
Studies with higher percentages of boys had, on average, higher effect sizes in the current study. This is in line with the idea that children who have more difficulties with EFs, such as boys compared to girls, are more likely to show improvements from experiences promoting EFs (Diamond & Lee, 2011). This is likely to do with the fact that there is a larger window for improvement. Unfortunately, there is, to our knowledge, no research available that systematically examined whether boys and girls are affected differently by interventions targeting EFs. A similar idea has been posed in research on the importance of teacher–child interactions for children’s development. The academic risk hypothesis suggests that at-risk children will benefit more from positive teacher–child interactions (Hamre & Pianta, 2001). A meta-analysis on the role of teacher–child relationships and engagement and academic achievement has shown similar results, with boys and low-SES children benefitting more from positive teacher–child relationships (Roorda et al., 2011). Based on this you would expect that our study would find higher effect sizes in low-SES samples, as low-SES children are considered to be at-risk and show in general lower EF performance (Hackman, Gallop, Evans, & Farah, 2015; Noble, Norman, & Farah, 2005). Results were, however, more mixed in this study. Between-study analysis showed that effect sizes were lower in low-SES samples. It is possible that for low-SES children the quality of the home environment is more important than the classroom context. On the other hand, it is important to note that this effect was not found within studies. A study that examined the teacher–child interactions in relation to EFs for a group of low- and high-SES children found that such a relationship was present only in the low-SES group. The reverse effect that was found between studies may have thus been due to other differences between the studies. Clearly, more research is needed that examines the association between teacher–child interactions and EFs in both low- and high-SES groups, while using the same design to draw more sound conclusions on this matter.
The question concerning which domains of teacher–child interactions affect which EFs could only partially be addressed due to the low number of studies available. Results show that teacher–child interactions are important for all EFs, though the quality of the dyadic teacher–child interactions seems to be more strongly associated with EFs than teacher–child interactions at the classroom level. It should be noted that although this moderator significantly predicted the effect size between studies, there was only a trend for this moderator within studies. Nevertheless, it seems important that prevention programs aimed at improving teacher–child interactions pay enough attention to the interactions with specific children and not only to the overall interactions (e.g., banking time, Playing2gether; Driscoll & Pianta, 2010; Vancraeyveldt, Verschueren, Van Craeyevelt, Wouters, & Colpin, 2015). Teachers should also attempt to spend enough time and energy in getting to know their students and the individual needs of the different students in order to maintain adequate relationships with each of them. However, this does not mean that the overall classroom climate and quality should be ignored. It should be noted that the difference found between these levels may be due to the measurements used. Dyadic relationships are generally assessed with teacher reports and sometimes with child reports, whereas classroom level interactions are generally measured with observations. Additionally, measurements of dyadic interactions, so far, focus only on the affective aspects of the relationship and do not include information on classroom organization and instructional support. As mentioned earlier, these two levels of interactions have been mainly investigated separately, and it is thus unclear whether classroom-level interactions may serve as a moderator in the association between dyadic interactions and EFs.
A final note of caution is warranted with regard to the fact that this meta-analysis examined potential moderating variables separately due to the limited number of studies. However, there were some associations found between moderator pairs, which may have influenced the results. First, higher effect sizes were found in samples with higher mean age and a higher percentage of boys, though there was a positive relationship between these moderators indicating that samples with a higher mean age also contained more boys. It is thus difficult to say which of these factors uniquely predicts the effect size. It is possible that gender came forward as a significant moderator only because samples with more boys also have a higher mean age. Second, the finding that teacher–child interactions were less related to EFs in low-SES samples may be driven by the level of teacher–child interactions that was examined. In low-SES samples it was mostly teacher–child interactions at the classroom level that was examined, and smaller effect sizes were found for these type of teacher–child relationships. Finally, higher effect sizes were found between teacher–child interactions at the dyadic level and EFs, though this level of interactions was examined in samples with older children. It is thus important that more research is being done, which would allow to examine more closely which moderators drive the effect between teacher–child interactions and EFs. This would give us valuable information about which teacher behaviors are important for EFs (e.g., emotional support or classroom organization) and for whom (e.g., boys or girls).
How Teacher–Child Interactions Might Affect EF
Although studies seem to suggest a positive association between teacher–child interactions and EF, little research provides us with explanations on why this is the case, or in other words what the underlying mechanisms are. Some suggestions can be found in the literature.
With regard to the affective domain (at both the classroom and the dyadic levels), two mechanisms might be involved in the effect of teacher–child interactions on EF. Several authors suggest that teacher–child interactions characterized by negative affectivity may have a negative effect on EFs because they form a source of stress for the child (Berry, 2012; Blair & Raver, 2015). Research has indeed shown that children who have more positive and less conflictual interactions with teachers have lower levels of salivary cortisol, an indication of lower stress levels (Hatfield, Hestenes, Kintner-Duffy, & O’Brien, 2013; Hatfield & Williford, 2017; Lisbonee, Mize, Payne, & Granger, 2008). A study of Ahnert et al. (2012) suggests that children involved in more positive teacher–child interactions not only have lower average stress levels but also can regulate their stress levels better, as is shown by reducing stress levels throughout the school day and week. In turn, high stress levels have been related to bad performance on EF tasks (e.g., Schoofs, Preuss, & Wolf, 2008). Some studies suggest that EF performance is optimal with medium levels of stress and worse when very low or high levels of stress arise (Lupien, Maheu, Tu, Fiocco, & Schramek, 2007). Research on parent–child interactions has shown evidence for this hypothesized pathway by showing that the effect of positive parenting on EFs is mediated by children’s cortisol levels (Blair et al., 2011).
Another potential mechanism through which positive affective interactions may promote EFs is through increased engagement and exploration (Cadima, Enrico, et al., 2016; Cadima, Verschueren, et al., 2016). When children are involved in positive teacher–child interactions, they will have more self-confidence and feel safe in their environment and are more likely to engage themselves in challenging activities in the classroom and persist in those activities when they become difficult (Cadima et al., 2015; Roorda et al., 2011). These children are thus more exposed to challenging tasks, which is likely to challenge their EFs and the use of such EFs in the classroom context. In other words, their EFs might develop better because they have more practice in the use of EFs through stimulating experiences. Such a pathway has been found in a study on parent–child interactions showing that children who are more attached to their parent will explore the home and classroom environment more and will perform better on a variety of cognitive tasks (O’Connor & McCartney, 2007). However, these tasks did not include EF tasks, and the sequence of these processes has, to our knowledge, not yet been examined for teacher–child interactions.
Several researchers assume that classroom organization should be most obviously linked to EFs (e.g., Downer et al., 2010), though the reasoning behind this is often not clarified. It is possible that children are less stressed in classrooms where it is clear what is expected from them and there are little interruptions and unpredictable situations during the classroom activities. However, an association between classroom organization and children’s stress levels has not yet, to our knowledge, been investigated. Another possibility posited by different researchers is that a well-organized classroom provides external help for children to organize their behavior and that this provides an ideal context for internalizing regulation strategies (Berry, 2012; Cadima, Enrico, et al., 2016; Cadima, Verschueren, et al., 2016; Choi et al., 2016; Rimm-Kaufman et al., 2009). Additionally, teachers who interact in a positive way and organize the classroom interactions and activities well are likely to show high levels of EFs themselves and as such may model the effective and efficient use of EFs in the classroom context (Choi et al., 2016; Rimm-Kaufman et al., 2009). Finally, research has shown that highly organized classrooms spend less time in transitions and more time in learning activities (Cameron, Connor, & Morrison, 2005). Children in highly organized classrooms may thus be more exposed to challenging learning activities during which they can practice their EF skills (Choi et al., 2016).
Finally, the domain of instructional support has been suggested to promote EFs in two ways. The first explanation refers to the fact that classrooms that score high on instructional support are characterized by activities that encourage reasoning and higher order thinking, such as classroom discussion and the use of open-ended questions (Downer et al., 2010). Such higher order thinking requires the use of EFs and thus might directly stimulate them (Cadima, Verschueren, et al., 2016; Fuhs et al., 2013). The second explanation refers to the importance of language in the development of EFs (Cadima, Verschueren, et al., 2016; Choi et al., 2016). Teachers who offer high levels of instructional support will facilitate language use and development (Downer et al., 2010). Children who have better language skills may use internal speech to guide their behavior more efficiently (Sarsour et al., 2011).
When and How EFs Are Studied in Relation to Teacher–Child Interactions
The current review clearly shows that research on the importance of teacher–child interactions for EFs has only emerged very recently and that the number of studies available are still limited. Most available studies focus on young children from preschool until the early years of primary education. This age range is important to focus on as it covers an important period during which both EFs and its underlying brain regions are characterized by rapid growth, change, and plasticity (Diamond, 2002; Klingberg, 2010; Kolb et al., 2012; Tsujimoto, 2008). As a consequence, stimulation from the environment is more likely to have a positive impact during this developmental stage (Huttenlocher, 2002). Nevertheless, it may be useful to explore the role of teacher–child interactions in other age periods as well—especially at the beginning of secondary education, when great changes in children’s environment occur at the same time when a second developmental spurt takes place (Zelazo & Carlson, 2012).
The current study also shows that the complexity of both the concept of teacher–child interactions and EFs (Nigg, 2016; Pianta, 2016; Verschueren & Koomen, 2012) results in fragmentation of the literature, making it difficult to draw overall conclusions. Moreover, some aspects of teacher–child interactions and EFs are more often examined, whereas others are ignored. With regard to EF, most studies assess behavioral forms of inhibition as a specific aspect of EF. Working memory, cognitive flexibility, and cognitive forms of inhibition are less well researched in relationship to teacher–child interactions. Especially in the case of working memory this provides a caveat, as this core EF has been linked to children’s learning-related behavior in the classroom (Fitzpatrick & Pagani, 2012; Gathercole et al., 2007) and has most consistently been related to children’s academic functioning (Cragg & Gilmore, 2014; Vandenbroucke, Verschueren, & Baeyens, 2017). Knowing which aspects of teacher–child interactions are important for this core EF can indicate which teacher behaviors should be promoted (e.g., teacher sensitivity vs. clarification of rules and expectations) to support working memory performance.
With regard to the teacher–child interactions this study found studies examining interactions at the dyadic level and at the classroom level in relation to EF. At the dyadic level focus is mainly placed on the quality of the affective interactions between the child and teacher. This is in line with the broader research examining the importance of teacher–child relationships for children’s development, where the attachment approach is most frequently used (Verschueren & Koomen, 2012). Nevertheless, other conceptual models on dyadic teacher–child relationships, such as a motivational or socialization models, may be useful in examining the role of the teacher EFs performance and development (Davis, 2003; Verschueren & Koomen, 2012). Additionally, within the attachment framework the concept of dependency in relation to EF was ignored (with the exception of Jones et al., 2013). One potential reason for the lack of research attending to the role of dependency is the discussion about the validity of the concept (Doumen et al., 2009; Roorda et al., 2011). Additionally, children seem to view high levels of dependency as a positive characteristic, whereas teachers generally interpret this dimension negatively (Vervoort et al., 2015). Nevertheless, more dependency generally relates to lower school adjustment, and it is interesting to examine how it relates to EFs (Murray & Murray, 2004). Children who are highly dependent on the teacher rarely explore their environment independently and, as such, have less learning experiences that can promote their development and EF. At the classroom level, there is a better balance between the different aspects of teacher–child interactions (emotional support, classroom organization, instructional support) researched in relation to EF.
Suggestions for Future Studies
The current study shows that there is already some evidence available that indicates the importance of positive teacher–child interactions for EF. However, additional research is needed to get more insight into what teachers can do or which teacher behaviors should be stimulated to promote children’s EF. We provide some directions to guide future researchers in this endeavor.
First, when examining the importance of teacher–child interactions for EFs close attention should be paid to the conceptualization of both concepts and the current scope should be broadened. More specifically, attention should be moved from behavioral forms of inhibition to working memory, cognitive forms of inhibition, and cognitive flexibility. Also, different types of tasks should be used to assure that findings are not task-specific (e.g., both verbal and visuospatial working memory tasks). Similarly, other aspects of teacher–child interactions should be considered at the dyadic level, including concepts such as dependency, autonomy support, and behavior management. Some of these factors have already been shown to be important for EFs when examined in parent–child interactions (e.g., Matte-Gagné & Bernier, 2011). Additionally, interactions between dyadic and classroom-level teacher–child interactions might be worth considering. For example, high levels of conflict might have a less negative effect on EFs when the general classroom climate is positive.
Second, it should be noted that the current study focused mainly on EFs as the cognitive aspect of children’s ability to regulate behavior, emotions, and thoughts (Zelazo & Carlson, 2012). However, research shows that cognitive and emotional aspects of self-regulation are both important and in daily life often intertwined (e.g., Ferrier, Bassett, & Denham, 2014). Some studies examining inhibition did use task that incorporate an emotional aspect (e.g., Toy Wrap). Emotional support from a teacher may have an even more profound effect on this emotional aspect of self-regulation. Future studies should therefore consider examining influences of children’s social environments on both cognitive and emotional aspects of self-regulation.
Third, it is important to note that all evidence reviewed in the current study is correlational in nature. None of the studies have attempted to manipulate aspects of teacher–child interactions through an experimental or intervention study design. A few studies did include an EF intervention, with a randomized controlled study design (e.g., Ursache et al., 2011). However, these interventions did not specifically focus on improving teacher–child interactions. Therefore, only the correlational data of these studies could be used in the overview and meta-analysis. As a consequence, we can only establish that there is a relationship between teacher–child interactions and children’s EFs, though it is unclear why this is and whether improving the interactions between children and their teachers would have a causal effect on EF. Future studies should attempt to manipulate aspects of teacher–student interactions. To this end, experimental studies could be used, as well as micro trials (Leijten et al., 2015) or single–case study designs (Smith, 2012). These last two types of research designs allow manipulation of very specific aspects of the interactions between the teacher and students (e.g., verbalizing experiences and feelings or sensitive responding), which will help us understand which specific behaviors of teachers might help improve EF performance and development in children. Additionally, these studies should pay attention to individual differences, examining for which children improving teacher–child interactions may benefit EFs (e.g., low- or high-SES children).
Conclusion
Overall, the current study attempted to provide an overview of the research examining the link between teacher–child interactions and children’s EFs and shows a positive association between both concepts, when measured both simultaneously and longitudinally. This suggests that teachers can promote the cognitive processes that are essential in children’s learning by changing their behavior to create an emotionally positive, structured, and cognitively stimulating classroom environment. It is, for example, important that teachers recognize children’s feelings and needs, adequately respond to these feelings and needs, make their expectations about children’s behavior clear, provide children with challenging classroom activities such as classroom discussions, and give extensive feedback about their behavior and performance. The dyadic relationships with teachers are especially important for children’s EFs. Furthermore, the current study suggests that teacher–child interactions are particularly important for EFs of children at the end of kindergarten and the beginning of primary education and for higher SES samples. Further research is needed to examine why this is the case and which specific teacher behaviors are important for children’s EFs and if there is a causal link between interactions with teachers and improvements in children’s EF performance. Such further insights can help building on to existing EF trainings, which are currently showing limited generalization to everyday situations (Diamond & Lee, 2011; Melby-Lervag & Hulme, 2013). Indirectly, our study suggests that preventing or intervening in EF problems might be more efficient when integrating such approaches in the classroom context (Hughes, 2011) and when including strategies that stimulate EFs both directly and indirectly (e.g., increasing children’s enjoyment and comfort by improving teacher sensitivity; Diamond, 2012).
Footnotes
Authors
LOREN VANDENBROUCKE is a PhD at the research unit of Parenting and Special Education, KU Leuven, Leuven, Belgium; email:
JANTINE SPILT is assistant professor at the research unit of School Psychology and Child and Adolescent Development, KU Leuven, Leuven, Belgium; email:
KARINE VERSCHUEREN is full professor at the research unit of School Psychology and Child and Adolescent Development, KU Leuven, Leuven, Belgium; email:
CLAIRE PICCININ is a master student of the Master of Psychology: Theory and Research, completing her internship at the research unit of Parenting and Special Education, KU Leuven, Leuven, Belgium; email:
DIETER BAEYENS is associate professor at the research unit of Parenting and Special Education, KU Leuven, Leuven, Belgium; email: