National and international research efforts in the field of early childhood education and care have increasingly examined child care quality as well as its effects on child development. The quality of the processes in the child care setting, especially the interactions between educator and child, is considered to be particularly important (Anders, 2013). There are numerous findings that show that the quality of educational processes is related to the cognitive, linguistic, and social development of children (Bäuerlein et al., 2013; Broberg et al., 1997; Côté et al., 2013; Duncan & National Institute of Child Health and Human Development Early Child Care Research Network, 2003; Tietze, 2013). As a result, increasing importance has been placed on the quality of teacher–child interactions. Interaction is closely related to the emergence of relationships and can thus have a significant influence on children's development. Previous studies have shown positive effects of high-quality teacher–child interaction on children's linguistic and cognitive abilities (Anders, 2013). Other aspects of child development could also be positively influenced by high-quality interaction. In addition to cognitive skills, social-emotional development can be a predictor of success at school and later at work. This includes the development of a stable personality as well as the establishment of healthy relationships with family and friends. Furthermore, there seems to be a connection between positive and balanced emotionality and the academic abilities of children (Son & Chang, 2018).

Social-Emotional Development

Many studies have shown that there is a close connection between emotional and social competence; therefore, both aspects have been included in the present research. Children’s relationships with their parents play a significant role in their social-emotional development. However, teachers in child care centers also are relevant interaction partners. Therefore, the reliability of caregivers, feelings of safety and security during first separation experiences are of fundamental importance to children (Gold & Dubowy, 2013). Since the domain of social-emotional development is rather broad, for the present research design, we focused on two aspects of this development: social competence and self-regulation.

Social Competence. Social competence is the ability to get into contact with others, to act independently in social contexts, to build relationships and friendships, to negotiate conflicts, and to adequately deal with one’s own feelings and needs (Gold & Dubowy, 2013). The age group of 22- to 45-month-old children examined in the present study may be considered too young to study for many of these aspects. For example, joint cooperative play is rarely observed until three to four years of age. Nevertheless, children have an early interest in peers and also play side by side, as well as with adults, before they are three years old (Viernickel, 2000).

Links between social competence and gender have been documented; for example, six-years-old boys seem to have less developed social skills than girls do (e.g. Ensor et al., 2011).

Self-Regulatory Skills. Self-regulatory skills as part of social-emotional competencies are essential for coping with cognitive and emotional tasks and challenges. Learning to reconcile one's own emotional experiences and related action tendencies with personal goals and social demands is a central developmental task of early childhood (Adrian et al., 2011).

Also, better emotional regulation skills are associated with better social-emotional competencies (Denham et al., 2012), more prosocial behavior, and better peer relations (Mostow et al., 2003). They also enable better adaptation to school and preschool environments and better school performance (Denham et al., 2012). For example, they help children to meet growing demands when they enter school, such as expectations regarding behavior, social skills, adaptability, teaching rules, sustained attention, or the control of emotional reactions (Son & Chang, 2018). In turn, various skills can be considered part of self-regulatory abilities: (1) Working memory, which is also crucial for successful learning, and is closely linked to executive functions. (2) Selective attention is important for self-regulation as is (3) inhibitory control (Vannice & Losoff, 2017). Previous studies have indicated that at the age of four to five years, boys perform more poorly than girls in tests of self-regulatory skills measured with the Head-Toes-Knees-Shoulders task and the Child Behavior Rating Scale (Matthews et al., 2009).

Relation Between Interaction Quality in Child Care Centers and Social-Emotional Development

Longitudinal studies have shown that only teacher–child interactions of particularly high quality may have a positive effect on early academic abilities and later school performance (Anders, 2013; Bäuerlein et al., 2013; Burchinal et al., 2010; Melhuish et al., 2015; Sylva et al., 2013). Compensatory effects for children from socially or culturally disadvantaged milieus, therefore, can only be expected in child care settings with high interaction quality (Burchinal et al., 2010; Magnuson et al., 2004; Watamura et al., 2011). For example, the social-emotional development of children with a migration background has been shown to be remarkably low in settings with low or medium quality of interactions (Beckh et al., 2014). In general, the quality of Early Childhood Education and Care (ECEC) in Germany and interaction quality as an essential part of it, has been determined in previous studies to be of medium level at most (Tietze, 2013). Medium to high qualities of emotional support and everyday organization have been reported in child care centers, while the area of learning support has generally been at a very low level (Tietze, 2013; von Suchodoletz et al., 2014; Wertfein et al., 2015).

The influence of the teacher–child interaction on the social-emotional development of children has increasingly become the subject of research efforts. However, there have been rather weak and partly contradictory results in the investigation of the connection between the experience of care (quality of teacher–child interaction) and its effects on children’s social-emotional development for children at about four years of age (Mashburn et al., 2008; NICHD Early Childcare Research Network, 2002; Pluess & Belsky, 2009). Some studies have shown that early child care has an impact on social-emotional development (Camilli et al., 2010). However, a study of the National Institute of Health and Disease (NICHD) of the USA found an association between many hours spent in child care and challenging behavior, regardless of child care quality (NICHD Early Child Care Research Network, 2002). Children in that longitudinal study were observed at 6 to 36 months and again when they were 54 months old. On the contrary, also based on the NICHD data, Pluess and Belsky (2009) found several associations with child care quality for both social behavior and problem behavior in children with difficult temperament. Their study observed both negative effects on children's development when the quality of child care was poor and positive effects when quality was good.

Son and Chang (2018) also analyzed the data from NICHD and focused on specific factors of social-emotional development; they provide evidence of relations between child care quality, including interaction quality, and children's self-regulatory skills. If children experience more positive interactions in a child care setting, they are more likely to score higher in sustained attention, in selective attention, and in self-regulation, in addition to being less impulsive. However, no relationship was found between child care quality and emotionality. Additionally, La Paro et al. (2014) measured behavior problems and socio-emotional competence in 11–41 months old children and found that interaction quality as measured with the CLASS-Toddler did not predict child social competence. In a study by Mashburn et al. (2008) with four-year-old children, only specific aspects of process quality showed effects on social-emotional development. No significant correlation was found between the general quality of the child care center and the development of social skills, nor between the quality of the instructional support (as measured with CLASS observation instrument) and the development of social skills. However, high-quality emotional interactions had positive effects on the development of children's social skills; a negative correlation between the quality of emotional interactions and problem behavior was also observed.

The British Effective Provision of Preschool Education (EPPE) study (Sylva et al., 2004) found that an overall high quality of child care is related to better social and behavioral development. The children were observed at age three to four years in child care and again when entering school, in the first and in the second grade. Indicators of quality included warm, interactive relationships between teachers and children. In addition, behavioral problems and antisocial behavior were observed to decrease if three-year-old children attended a facility that provided high quality child care. A relevant study examining the relationship between experiences of child care and social-emotional development at 36 months, on the other hand, did not find any effect of child care quality on disruptive behavior (Barnes et al., 2010).

The German national study on early childhood education and care (NUBBEK; Tietze, 2013) also found only partial associations between child care quality and social-emotional development. For four-year-old children, higher general process quality scores were associated with higher levels of social competence and less problem behavior, when children were assessed by mothers. However, teacher–child interaction quality was also evaluated using the Caregiver Interaction Scale (Arnett, 1989) with the categories of sensitivity, involvement, and acceptance. Higher values for sensitive interaction were related to lower scores for social competence and higher scores for problem behavior, according to the assessments of mothers. Again, no corresponding correlation was found for two-year-old children. Conversely, if the quality of the teacher–child relationship was assessed by the teacher using the Pianta scale (Pianta, 1994), positive effects for social-emotional development and problem behavior were found for both two- and four-year-old children (Tietze, 2013). These results highlight the importance of measurement questions for the topic. In a recent study in Finland and Portugal with similarly aged children (Salminen et al., 2021), researchers found for Finland that teachers’ engaged support for learning was positively associated with children's selective attention and inhibitory control, while teachers’ emotional and behavioral support was positively associated with children's inhibitory control. For Portugal, engaged support for learning was positively associated with children's selective attention.

It should be kept in mind that in each of the aforementioned studies, different aspects of (1) interaction quality and (2) social-emotional skills were studied with different instruments in different societies and settings. Moreover, as expected, the effects of interaction quality are minimal because influences on child development are very diverse (Anders, 2013; NICHD Early Child Care Research Network, 2002). Therefore, there is a need for research on the relationship between interaction quality and child outcomes, especially for very young children.

The Present Study

In the present research project, the extent to which there is a relationship between the quality of teacher–child interactions in early childhood institutional care and children's social-emotional development was investigated. For this purpose, an observational study was conducted with a cross-sectional design. Interaction quality was assessed by means of video observation with the CLASS Toddler instrument, while self-regulatory and social-emotional skills were surveyed by means of standardized tests and teacher assessments. Based on the previous literature, we stated the following research hypotheses:


We hypothesized that higher quality of teacher–child interaction does associate with higher scores in children’s self-regulatory skills, as measured by selective attention, working memory and inhibitory control.


We also hypothesized that higher quality of teacher–child interaction does associate with higher scores in children’s self-regulation and social skills rated by their teachers as measured by the MASCS and the CBRS.


According to previous findings, we expect girls to obtain higher values in the included aspects of self-regulatory skills and social skills (Ensor et al., 2011; Junttila et al., 2006; Matthews et al., 2009).


The present study was part of a larger collaborative research project conducted in conjunction with the University of Jyväskylä, Finland (see e.g., Salminen et al., 2021). The study methods were coordinated to a significant extent to enable intercultural comparability. As part of the project, various developmental domains were measured: mathematical skills, literacy and language skills, as well as social-emotional skills. The overall research project additionally focused on the effects of professional knowledge and beliefs, leadership quality, and peer interactions, as well as the home learning environment. The study conforms with recognized standards (e.g., Declaration of Helsinki) and was approved by an internal ethics committee at the University of Applied Sciences, Potsdam. The parents of participants gave informed consent for their children to participate in this study.


The study involved children (N = 64; 44% girls) from 22 to 45 months of age who attended a child care center in a medium-sized eastern German city. Additionally, six children participated in the study but were excluded because they did not finish more than one task. The children and teachers participating in the study were part of the child care center for at least three months. For children to be included in the sample, they had to have a sufficient grasp of the German language to be able to understand the instructions of the test leader and implement these instructions. Each child received a certificate at the end of the session. Payment was not included.

To measure the quality of teacher–child interaction, the group teachers (N = 9) and at least five children from 22 to 45 months in age of each group were included. Only group teachers where at least five parents gave informed consent were included in the study. Observation material from nine professionals in nine groups was available. The professionals, all female, were on average 36.89 years old (SD = 7.85), had on average 8.59 years (SD = 5.69) of professional experience as a pedagogical professional, and had been working in the facility for an average of 5.1 years (SD = 3.27). All participants had completed training for pedagogical work with children, of which 90% had completed a professional apprenticeship as a state-certified educator (N = 8) and one person has completed a master's degree in pedagogy. Regarding the educational background of the families, only a small proportion of families had no vocational training (n = 3). Forty-eight percent of parents had completed or were in vocational training. Almost half of the parents had completed higher education.


Assessment of Interaction Quality. The CLASS Toddler (La Paro et al., 2012) was used to assess interaction quality. CLASS Toddler is used to measure process quality at the group level. It covers the age range of 15 to 35 months. It is used to evaluate the quality of interactions between the professional and the child regarding two dimensions: (1) ‘emotional and behavioral support’, including the items positive climate, negative climate, teacher sensitivity, regard for child perspectives, and behavior guidance; and (2) ‘engaged support for learning’, including the items facilitation of learning and development, quality of feedback, and language modeling. Four different types of activities, including play, care routine, educational or emerging academic activities, and creative activities have to be recorded. In each of the nine groups the relevant teacher was video-recorded while interacting with children in the respective four different situations. Therefore, observation material is available for a total of 36 sequences. Quality is assessed on a seven-point scale in which one to two points represent low quality, three to five represent medium quality, and six to seven represent high quality.

A peculiarity of the CLASS Toddler instrument is the special way of scoring prescribed in the manual and trainings to achieve the user license: In a first step, the observation material is rated based on the multiple items of each dimension. But in the second step, these ratings are not summed up but disregarded and a second, independent rating is provided for the whole dimension as a single score. We wondered what effect this special way of measuring might have on the results and present in the analysis both ways of measuring in parallel for the sake of comparison.

Please note also that CLASS Toddler was developed to cover the age range of 15 to 35 month old children. It is an instrument that measures the quality of the teacher who interacts with children. Although we included some children older than 35 months, the interaction quality was observed for whole classrooms and not for specific children. Raters did not face any difficulty in rating the video sequences. Other studies used CLASS Toddler for similar age ranges (Salminen et al., 2021).

Self-Regulation Skills. Three aspects of self-regulation are examined in individual assessments with the children:

Working Memory

To evaluate working memory, the “Hidden Toys Task” (Mulder et al., 2014) was used. Six boxes with lids are placed in front of the child. Children are shown how toy animals are hidden in the boxes, with one in each box. Children are then asked to find one of the animals. The test leader then takes out the animal and closes the box before children are asked to find another toy. Between each searching exercise, children are distracted for six seconds. For the working memory analysis, the number of animals found was used. Since there were six boxes with one animal each, a maximum of 6 points could be scored.

Selective Attention

Selective attention was evaluated using one subtest of the neuropsychological assessment (NEPSY; Korkman et al., 1998). Children were given a sheet of paper with many small pictures of animals, humans, toys, etc. Rabbits are depicted in one of the pictures, and children were asked to find and put a stamp on it as quickly as possible. The same procedure is carried out on the back of the sheet where children are instructed to identify cats. The time limit is three minutes for each side. During the task, time is tracked and the number of correct items (maximum 20 rabbits, maximum 20 cats) and incorrect items is counted.

Inhibitory Control

The children's ability of inhibitory control was assessed by using the Toy Wrap task (Smith-Donald et al., 2007). Children are told that the test leader has a surprise that the test leader must first gift wrap. In the meantime, children are not allowed to peek. After 60 s children are allowed to look. The time until children peek is calculated. Inhibitory Control was operationalized by the duration (in seconds) children waited until they looked for the first time after the wrapped toy, even though they were asked not to look. According to other studies, if the child did not look, the maximum time of 60 s was used (Salminen et al., 2021).

Teacher Rating

Additionally, the children's self-regulation is evaluated by their teachers using a shortened version of the Child Behavior Rating Scale that contains 17 statements (Bronson et al., 1990); here, teachers must give a rating from one (never) to five (always) of how often a particular behavior or a similar situation occurs in the day-to-day operations of the child care center. The aim is to assess how children can control their own attention, emotions, and behavior and how they can adapt to the demands of the child care center. This includes how the child handles equipment and assignments within the institution.

Social-emotional Skills. Children’s social skills are measured using teacher ratings with the Multi-Source Assessment of Children's Social Competence (MASCS; Junttila et al., 2006). This assessment contains the two main dimensions of prosocial behavior (cooperative skills, empathy) and antisocial behavior (impulsiveness, and disruptive behavior). Each child should be rated by the teacher regarding on a scale of one (never) to four (very often) with regard to 13 statements on a scale from 1 to 4.


Observation and assessment took place in nine child care groups on a typical morning at child care centers over the course of five hours. Two trained persons visited the child care center on two consecutive days. Four sequences per group were recorded on video, including creative activities, educational activities, care routines, and play, with a minimum sequence length of 10 min and a maximum of 20 min. All tests were administered by trained test administrators and lasted approximately 20–30 min. Appropriate child-adapted breaks were implemented. If children could not perform actions on a task or signaled stress, the test was stopped to prevent feelings of frustration. This occurred for six children. The pedagogical staff received a briefing on the process and procedure of video recording. They were informed beforehand that four different settings would be used to assess the quality of interaction. They were not given any special instructions on how to behave; they were only told to organize their activities just as usual. The children were told that everything they did and experienced in the kindergarten would be filmed that day.


To ensure the quality of the observation ratings in areas such as objectivity and reliability, all three observers completed basic training on the instruments by a trainer certified by Teachstone. The three observers completed the certification with a reliability of 93% each, achieving gold standard according to the manual. The interrater reliability of the sample for the 20% double-coded observations from the total material was very good (ICC = 0.96). Observer agreement was determined according to guidelines found in the manual and following other studies (La Paro et al., 2012), with ± 1 scale value considered agreement in the codification.



Structural equation modeling (SEM) methodology was adopted in R (Rosseel, 2012), to model the relations between interaction quality, self-regulation and social-emotional competence. Broadly speaking, SEMs combine a measurement model, as obtained with a common factor analysis, with a structural model that reflects the relations between the to-be-investigated latent constructs similar to a common linear regression. This approach is favorable in terms of reducing measurement error compared to a common regression analysis based on sum or average scores.

As regards the measurement model, interaction quality was considered to be a construct composed of the dimensions ‘positive climate’, ‘negative climate’, ‘teacher sensitivity’, ‘regard for child perspectives’, and ‘behavioral guidance’. All five dimensions represent the higher-level domain ‘emotional and behavioral support’ of the CLASS Toddler scale. We chose to include only this domain because it seemed most relevant for the present purposes.

Regarding the construct ‘self-regulation’, we conceptually distinguished between (a) tasks used to measure children’s self-regulatory capacities and (b) teacher’s perceptions of children’s self-regulation. Dimensions of the former were ‘working memory’, ‘selective attention’, and ‘inhibitory control’, whereas dimensions for the latter were ‘classroom’ and ‘interpersonal’ self-regulation.

On a technical note, we would like to point out that although the data included multiple levels (nested observations) which could explicitly be modeled, we refrained from doing so, because our sample size was too small. In the following, we will report separate analyses for the multiple-item measure and the single-scoring measure of interaction quality with the Class Toddler instrument to compare both scoring approaches.

Descriptive Statistics

Interaction Quality (multiple-item-based analysis). Means, standard deviations, minimum and maximum values, as well as skew and kurtosis were computed to investigate the response characteristics of each dimension of interaction quality. As can be seen in Table 1, most indicators seemed to be properly distributed suggesting that the measurement of interaction quality was overall successful. However, the dimension ‘negative climate’ exhibited signs of ceiling effects for all indicators as indicated by a high mean, low standard deviation and extreme skew and kurtosis values. This being confirmed by a visual inspection of the data, we decided to remove this dimension from further analyses. For ‘behavioral guidance’, the indicator ‘problem behavior’ likewise showed extreme skew and kurtosis values, and visual inspection, again, suggested a ceiling effect, so we decided to remove the item ‘Problem behavior’ from the dimension.

Table 1 Descriptive statistics for interaction quality

To investigate whether the conceptual five-dimensional structure was reflected in the remaining data, we conducted an exploratory factor analysis with the number of factors being determined using a scree plot. As a result, a single dimension was suggested, and all items loaded on this factor with loadings higher than 0.5. Confirming this outcome, Cronbach’s Alpha value was estimated at α = 0.95 (very good). Taken together, this pre-analysis suggested that all items indeed measure a single construct (interaction quality), but there is not enough reason to separate this construct further into different dimensions. We therefore adopted the one-factor model and subsequently summarized the values of each indicator per teacher to obtain a per-teacher score that could be used for the later regression model.Footnote 1

Interaction quality (single-score analysis). In accordance with the multiple-item-based analysis, the dimension ‘negative climate’ exhibited signs of a ceiling effect (Table 1, lower part), as its mean was close to the maximum (M = 6.69, SD = 0.53) and its distribution considerably skewed (skew = − 1.37). For this reason, we chose to remove this dimension. In contrast to the multiple-item-based analysis, the dimension ‘behavioral guidance’ seemed to be properly distributed. To foreshadow the results, however, including this dimension into the regression models resulted in severe problems with model fit and, as a result, we removed this dimension completely from the analysis. To investigate whether the remaining three indicators formed separate dimensions or, as was the case with the multiple-item-based analysis, a single dimension, we conducted an exploratory factor analysis with the number of factors being determined using a scree plot. As a result, a single dimension was suggested, in accordance with the multiple-item-based analysis. Cronbach’s Alpha was estimated at α = 0.91(very good) suggesting a high internal consistency for this single-factor model.

Self-regulation (tasks). Self-regulation, as measured by behavioral tasks, consisted of the dimensions ‘working memory’, ‘inhibitory control’, and ‘selective attention’ (see Table 2). Inspection of the working memory measure revealed an appropriate difficulty of this task (M = 4.46, SD = 1.18), but also an obvious deviation from a normal distribution (skew = − 1.42, kurtosis = 2.69). Because we did not want to drop this measure altogether, we decided to apply a square power transformation on this measure, which alleviated these problems (skew = − 0.34, kurtosis = − 0.28). As our measure of selective attention, we scored the respective subtest of NEPSY and an overall score was calculated based on processing time and percentage of correct and incorrect reactions. This measure was of appropriate difficulty, as well as reasonably distributed (M = 8.67, SD = 1.75, skew = 0.47, kurtosis = − 0.16) and no transformation needed to be applied. Regarding the ‘inhibitory control’ measure, it was evident that 16 of the 64 children succeeded in not peeking at all at the wrapped toy as reflected by the number of seconds waiting being at the ceiling (60 s). The majority of children, however, seemingly could not wait longer than 20 s (N = 35), and the remaining children gave up halfway (N = 12). Since there did not seem to be an appropriate transformation that solves both the ceiling effect and the non-normal distribution, we decided to leave the measure as it was and to keep this problem in mind during the subsequent analyses. Finally, we checked whether the three tasks reflected a single construct of self-regulation by computing the correlations between them. It was evident, however, that performance in the three tasks only weakly correlated (r = [0.14; 0.27]), meaning that combining the tasks into a single variable was not the right choice. We therefore analyzed the effect of interaction quality on three separate facets of self-regulation.

Table 2 Descriptive Statistics for Self-Regulation

Self-regulation (ratings). Regarding the teacher ratings of self-regulation (Table 2), all items of the shortened CBRS showed proper distributional characteristics. That is, item means were distributed between 2.80 (min) and 3.66 (max), standard deviations between 0.84 (min) and 1.26 (max), and both skew and kurtosis values were below the critical value of |1| for all items. We subsequently checked the factor structure of this questionnaire using exploratory factor analysis. As a result, a two-factor solution was preferred, in accordance with the conceptual distinction between classroom and interpersonal self-regulation. That being said, four items from the classroom dimension and two items from the interpersonal dimension did not load sufficiently on any factor (< 0.5) meaning that we had to drop these items from further analyses (see Table 2). For the remaining items, Cronbach’s Alpha was estimated at 0.89 and 0.88, respectively.

Social-emotional Skills. The MASCS questionnaire exhibited proper distributional characteristics (Table 3). In particular, item means were distributed between 1.86 (min) and 3.25 (max), standard deviations between 0.65 (min) and 1.06 (max), and both skew and kurtosis values were below the critical value of |1| for all items. We therefore kept all items for this scale. We further checked whether the two conceptual dimensions inherent in this scale (‘prosocial behavior’ and ‘antisocial behavior’) were reflected in the data at hand by conducting an exploratory factor analysis. As a result, a two-factor solution was indeed preferred, and all items loaded on their ‘correct’ dimensions having a loading of at least 0.5. Consistent with this result, Cronbach’s Alpha for each dimension were estimated at 0.84 (prosocial) and 0.89 (antisocial), respectively.

Table 3 Descriptive statistics for social-emotional skills

Behavioral Self-Regulation Tasks and Interaction Quality

We subsequently regressed each task measure of self-regulation on interaction quality, age and gender being included as control variables. In the following, we report standardized regression coefficients (labeled ‘b’) and confidence intervals (labeled ‘CI’). In case of marginal significance, the confidence level of the CI is 90%, otherwise, it is 95%. Further note for the regressions of the multiple-item-based interaction quality score on the task measures of self-regulation: These models were just-identified meaning that it did not make sense to report fit criteria (fit is always perfect in these cases).

Multiple-item-based interaction quality. For ‘working memory’, the resulting model revealed effects in the expected direction; that is, higher interaction quality scores were associated with higher scores in the working memory task (b = 0.23; CI = [0.01; 0.45]). Second, there was a significant effect of gender (b = − 0.26; CI = [− 0.48; − 0.05]) suggesting that being male was associated with lower scores. Finally, age was positively associated with higher scores (b = 0.36; CI = [0.15; 0.57]). Regarding our measure of ‘selective attention’, there were no significant effects for interaction quality (b = − 0.01; CI = [− 0.26; 0.23]) and gender (b = − 0.19; CI = [-0.42; 0.04]). However, there was a significant positive effect of age (b = 0.34; CI = [0.11; 0.56]) suggesting that older children showed better selective attention. The adjusted R2 was estimated at 0.10. For ‘inhibitory control’, there was no significant effect for interaction quality (b = 0.10; CI = [− 0.13; 0.33]), whereas the control variable gender was marginally significant (b = − 0.21; CI = [− 0.39; − 0.02]). There was a significant positive effect of age (b = 0.40; CI = [0.20; 0.60]) suggesting that older children showed better inhibitory control. One must keep in mind, however, that our measure of inhibitory control contained a ceiling effect and non-proper distributional characteristics, meaning that one needs to be cautious with regard to interpreting this result.

Single-score interaction quality. The corresponding regressions for our single-score measure of interaction quality generally revealed analogous effects. For ‘working memory’, there were significant effects of interaction quality on the working memory task (b = 0.27; CI = [0.06; 0.49]) and of gender (b = − 0.28; CI = [− 0.49; − 0.07]) and age (b = 0.33; CI = [0.13; 0.54]). One must note, however, that model fit was not ideal (χ2 = 37.04, p > 0.05, df = 8, CFI = 0.90, TLI = 0.830, RMSEA = 0.25) meaning that results must interpreted with caution.

Regarding ‘selective attention’, there were again no significant effects of interaction quality (b = 0.01; CI = [− 0.22; 0.25]) and the control variable gender (b = − 0.20; CI = [− 0.42; 0.03]). The significant positive effect of age, however, was present also in this analysis (b = 0.33; CI = [0.11; 0.54]). Model fit was slightly better although, again, not ideal (χ= 29.49, p < 0.05, df = 8, CFI = 0.92, TLI = 0.87, RMSEA = 0.21). For ‘inhibitory control’, there was no significant effect for interaction quality (b = 0.17; CI = [− 0.05; 0.39]), whereas the control variable gender reached the 5% significance level (b = − 0.22; CI = [− 0.43; − 0.01]). In addition, there was a significant positive effect of age (b = 0.38; CI = [0.18; 0.57]) suggesting that older children showed better inhibitory control. Model fit was comparable to the other two models (χ= 33.45, p < 0.05, df = 8, CFI = 0.92, TLI = 0.85, RMSEA = 0.22).

Self-regulation Ratings and Interaction Quality

The ‘classroom’ and the ‘interpersonal’ dimension of the CBRS were modeled as latent variables and separately regressed on interaction quality, age and gender being additional control variables.

Multiple-item-based interaction quality. The first model with the ‘classroom’ dimension exhibited a very good fit, as indicated by common fit criteria for structural equation models (Kline, 2011): χ= 25.59, p > 0.05, df = 24, CFI = 0.997, TLI = 0.998, RMSEA = 0.04. However, the effect of interaction quality on classroom self-regulation was non-significant (b = − 0.07; CI = [− 0.42; 0.28]), and so was the effect of age (b = − 0.06; CI = [− 0.41; 0.30]). Gender, however, showed a significant association with self-regulation (b = − 0.42; CI = [− 0.72; − 0.12]) suggesting that being male was associated with lower classroom self-regulation scores. For the ‘interpersonal’ dimension, model fit was similarly good (χ= 21.34, p > 0.05, df = 17, CFI = 0.99, TLI = 0.996, RMSEA = 0.07), but again, there was no significant relation between interaction quality and self-regulation (b = 0.07; CI = [− 0.21; − 0.35]). Here, age showed a significant effect (b = 0.35; CI = [0.10; 0.59]) suggesting that older children showed better interpersonal self-regulation. The effect of gender, however, was non-significant (b = − 0.14; CI = [− 0.40; 0.13]).

Single-score interaction quality. Comparable results were obtained for the single-score analysis. Both models of self-regulation exhibited acceptable fits (classroom: χ= 52.69, p < 0.05, df = 33, CFI = 0.94, TLI = 0.95, RMSEA = 0.11; interpersonal: χ= 51.26, p < 0.05, df = 33, CFI = 0.96, TLI = 0.97, RMSEA = 0.098). For classroom self-regulation, there was an effect of gender (b = − 0.44; CI = [− 0.69; − 0.18]), but non-significant effects for interaction quality (b = − 0.10; CI = [− 0.31; 0.10]) and age (b = 0.03; CI = [− 0.27; 0.34]). For interpersonal self-regulation, there was a significant effect of age (b = 0.36; CI = [0.12; 0.60]), but non-significant effects of interaction quality (b = 0.09; CI = [− 0.10; 0.28]) and gender (b = − 0.13; CI = [− 0.39; 0.13]).

Social-emotional Skills and Interaction Quality

Multiple-item-based interaction quality. In accordance with the two-factor structure of the MASCS reported above, ‘prosocial behavior’ and ‘antisocial behavior’ were separately regressed on interaction quality, age and gender being included as control variables. As a result, both regression models exhibited acceptable fits (prosocial: χ= 54.24, p < 0.05, df = 41, CFI = 0.95, TLI = 0.96, RMSEA = 0.08; antisocial: χ= 57.01, p < 0.05, df = 32, CFI = 0.97, TLI = 0.98, RMSEA = 0.12). However, although pointing into the expected direction, the effect of interaction quality remained non-significant in both cases (prosocial: b = 0.03; CI = [− 0.24; 0.31]; antisocial: b = − 0.18; CI = [− 0.47; 0.11]). For prosocial behavior, there were significant effects of gender (b = − 0.42; CI = [− 0.64; − 0.20]) and age (b = 0.35; CI = [0.10; 0.59]) suggesting that this kind of behavior is more prevalent in girls than in boys and that it becomes more prevalent with age. For antisocial behavior, both effects remained non-significant (gender: b = 0.12; CI = [− 0.17; 0.40]; age: b = − 0.16; CI = [− 0.43; 0.11]).

Single-score interaction quality. In contrast to the multiple-item-based analysis, the model for prosocial behavior exhibited a relatively poor fit (χ= 74.98, p < 0.05, df = 63, CFI = 0.881, TLI = 0.897, RMSEA = 0.090). Although pointing into the right direction, the effect of interaction quality on prosocial behavior failed to reach significance (b = 0.16; CI = [− 0.06; 0.38]). However, there were significant effects of gender (b = − 0.43; CI = [− 0.66; − 0.22]) and age (b = 0.31; CI = [0.03; 0.59]), consistent with the multiple-item-based analysis. The regression modeling the relation between interaction quality and antisocial behavior exhibited an acceptable fit (χ= 87.92, p < 0.05, df = 52, CFI = 0.94, TLI = 0.945, RMSEA = 0.114). In contrast to the multiple-item-based analysis, there was a marginally significant effect of single-score interaction quality on antisocial behavior (b = − 0.20; CI = [− 0.39; − 0.02]) in the expected direction; that is, higher interaction quality scores were associated with fewer antisocial behavior. The effects of gender (b = 0.06; CI = [− 0.21; 0.35]) and age (b = − 0.20; CI = [− 0.48; 0.08]), however, were non-significant. To explore in more detail the relation between interaction quality and antisocial behavior, we carried out additional regressions separately for each sub-factor of antisocial behavior, that is, disruptiveness and impulsiveness. As a result, there was a significant effect of interaction quality on disruptiveness (b = − 0.29; CI = [− 0.8; − 0.09]) but not on impulsiveness (b = − 0.11; CI = [− 0.38; 0.15]) suggesting that higher interaction quality may be associated with lower disruptiveness. In addition, there was a significant effect of gender on impulsiveness (b = 0.28; CI = [0.02; 0.5]) but not on disruptiveness (b = − 0.06; CI = [− 0.33; 0.24]). Age did not significantly relate to either dimension of antisocial behavior (disruptiveness: b = − 0.23; CI = [− 0.50; 0.05]; impulsiveness: b = − 0.22; CI = [− 0.48; 0.04]).


How does teacher–child interaction quality in early childhood care and education settings influence the development of social-emotional skills in children around three years of age? We found evidence in an observation study supporting the hypothesis that even for children this young, the quality of interaction in child care settings is positively related to working memory and less disruptive behavior. Conversely, we found no evidence that interaction quality influenced other aspects of social-emotional development, such as selective attention, inhibitory control, classroom and interpersonal self-regulatory skills, prosocial behavior, and impulsiveness. Additionally, age proved to be a strong predictor of all the cognitive-oriented dimensions (i.e., selective attention, inhibitory control, and working memory), as well as for interpersonal self-regulation and prosocial behavior. Gender had an influence in the expected direction on working memory, inhibitory control, classroom self-regulation, and prosocial behavior with girls scoring higher than boys in each of these domains.

Theoretical Implications

These observations have important implications for both theory and practice. From a theoretical perspective, our results provide further support for the assumption that the quality of early education settings does matter for children’s early development—especially in the domain of self-regulation and social-emotional skills. The results are thus in line with previous studies that found positive relationships between child care quality and social-emotional skill development (Camilli et al., 2010; Mashburn et al., 2008; Sylva et al., 2004).

Children in groups with higher interaction quality are more likely to have better working memories, as well as show less disruptive behavior. These findings correspond to those of Son and Chang (2018) but they relativize the null findings of Barnes et al. (2010). Our findings confirm the assumption that interaction quality is of high importance for children’s development, even at a very early age and after low dosage of child care experience. However, an inverse relationship is also possible: Children with better self-regulation skills and better social behavior might experience more favorable interactions than children who show fewer self-regulation skills or more adverse social behavior. Research examining individual children and their individual experiences with child care quality is needed to address this question.

It is also interesting to discuss the null findings: Why is interaction quality related to working memory but not selective attention or inhibitory control, for example? In the study by Salminen and colleagues (2021) with children of a similar age in Finland and Portugal, selective attention only correlated with the dimension ‘Engaged Support for Learning’ but not with the dimension ‘Behavioral and Emotional Support’ (Salminen et al., 2021). This might support our finding that the dimension ‘Behavioral and Emotional Support’ is not related to selective attention in our study. Possible reasons for null findings are various: First, it might be that effects only occur after a longer period of attending child care. Second, this could be due to the small sample size. Third, the influences on child development are manifold, and family background variables presumably play an important role as well (Anders, 2013; Gold & Dubowy, 2013). Indeed, the NUBBEK-study (Tietze et al., 2013) found that the effect of family background variables is four times stronger than the effect of child care quality. Another possibility is that interaction quality was not good enough to show an effect on child outcomes. As Burchinal and colleagues (2010) analyzed, there is a threshold for interaction quality below which associations are less likely to be seen.

The effect of age is as expected since development progresses quickly in this early age, specifically in the area of self-regulation skills. This explains the age effects we found for selective attention, inhibitory control, and working memory, as well as interpersonal self-regulation. Our findings regarding gender effects are also in line with previous findings (Ensor et al., 2011; Junttila et al., 2006; Matthews et al., 2009).

Methodological Implications

For the CLASS Toddler measure, we compared the results of the prescribed single-score measure with a measure based on the multiple item ratings. Both kinds of rating are provided in the course of applying the instrument by default. However, only the single-score measure is used in the analyses. From a statistical perspective, the multiple-item-based measure might be more robust and should be preferred. Therefore, in the present study, we compared the results of both ways of measurement. Both ways of measuring teacher–child interaction quality yielded mostly similar results. Thus, it seems justified to stick to the prescribed way of giving single scores for each dimension as laid out in the manual of the CLASS Toddler instrument.

Limitations of the study

First of all, sample size of the present study is rather small due to organizational challenges of such a comprehensive study. This leads to low statistical power. Bigger samples and longitudinal designs are needed to verify the results. Regarding the procedure, in standardized test situations for self-regulatory skills, we observed that it was difficult for many children of this age to concentrate for more than a few minutes. As a result, varying numbers of breaks had to be administered during the testing session, which may have influenced the measurement results. Teachers were informed that their interactions with children were under observation. This may have impacted their behavior during the observations. However, participants have to give informed consent to their being video-recorded and studied. Data show sufficient variability and are far from ceiling effects as is shown by the low-quality results in the domain of engaged support for learning. Another caveat might be the mixed method approach, whereby the same participants that are assessed also provide measurement data for the children. The reason for this design was that testing capacity with young children is very limited. The study contained a high number of direct assessments for the children (including early math and literacy development). Therefore, a teacher rating is another valid source of data about child outcomes; as is also demonstrated in the recent studies by Staff and colleagues (2021) and Senol and Turan (2019).

Overall, the age group to be examined is rather young. Thus, especially with the younger children who were around their second birthday, it was not always possible to assess with certainty to what extent they completely understood the task. The Toy Wrap task and NEPSY were developed for children older than three. In the Hidden Toys task, for example, “animals” are mentioned first and “toys” later. In some instances, when asked to find a "toy," children got up to get a toy from somewhere else and did not look in one of the boxes in front of them. However, for lack of a better alternative, these instruments were used in this and previous studies. In general, it is difficult to find instruments to measure social-emotional and self-regulatory child-outcomes that are suitable for toddlers (La Paro et al., 2014).

In addition, it is important to note that while CLASS Toddler is used internationally in studies and thus allows for comparability, the instrument was developed in the USA and is based on a culture-specific view of quality, which is not equivalent to what is considered to be good quality in German society (Mayer & Beckh, 2018). We chose to use the CLASS Toddler in the present study nevertheless, since it has been established as the major instrument to measure process quality internationally. Other well-known instruments, like the Caregiver Interaction Scale (Arnett, 1989) or the Student–Teacher Relationship Scale (Pianta, 2001) do not include sub-scales regarding learning support. Additionally, according to Colwell (2013) the CIS is not apt to distinguish between highly vs. moderately positive interactions. The STRS is a self-report measure assessing teachers’ perception of the quality of their relationship with students. Thus it is a more subjective measurement than the CLASS Toddler, which is based on video-recording and observation.

Practical Implications

The abovementioned results also have consequences for practice, such as indicating the need for higher quality standards, especially regarding interaction quality (Mashburn et al., 2008). If there is an effect for young children aged two to four years, it is to be expected that interaction quality has a great importance for children`s development. On the other hand, there is a need for further research, especially for children under the age of three, to gain a more specific understanding of the influences of institutionalized child care settings on early development.


Despite the limitations described, valuable insights and implications for further research as well as for educational practice can be derived. The results indicate that even after only a short period of institutional child care, teacher–child interaction quality can impact children’s early social-emotional development. The present findings reinforce the need for instruments measuring self-regulation and more differentiated studies, especially for children under three years of age.