Who benefits from teaching quality: the role of prior achievement in Czech lower secondary schools

Sedova, Klara; Sedlacek, Martin

doi:10.3389/feduc.2025.1712000

ORIGINAL RESEARCH article

Front. Educ., 20 January 2026

Sec. Assessment, Testing and Applied Measurement

Volume 10 - 2025 | https://doi.org/10.3389/feduc.2025.1712000

Who benefits from teaching quality: the role of prior achievement in Czech lower secondary schools

Klara Sedova^*

Martin Sedlacek

Faculty of Arts, Department of Educational Sciences, Masaryk University, Brno, Czechia

Teaching quality is widely recognized as a crucial factor influencing student achievement. The Three Basic Dimensions (TBD) framework—comprising cognitive activation, classroom management, and student support—is commonly used to assess instructional quality, but its validity and effects remain not well understood. This study examined the factorial structure of the TBD model in Czech lower secondary schools and explored how dimensions of teaching quality predict mathematics achievement, with a focus on differences across students’ prior achievement. The sample included 2,609 seventh-grade students from 142 classrooms in Czech lower secondary schools. Student ratings of teaching quality were used to test the factorial structure of the TBD model via confirmatory factor analysis. A four-factor model was tested, distinguishing between cognitive activation through learning tasks (CAT) and through teacher interaction (CAI). Multilevel structural equation modeling was applied to assess the effects of teaching quality on mathematics achievement while controlling for socioeconomic status, gender, and prior achievement. The four-factor model provided a better fit than the original three-factor structure. CAI at the individual level and classroom management at the class level significantly predicted student achievement. However, these effects were not uniform: while high-achieving students benefited from CAI, and moderately achieving students from classroom management, lower-achieving students did not benefit from any of the observed teaching quality dimensions. The study highlights that teaching quality is not universally beneficial for all students. Instructional practices that are generally effective may fail to support lower-achieving students, potentially reinforcing existing achievement gaps.

1 Introduction

Teaching quality, a crucial factor influencing student learning outcomes, has been extensively researched through various frameworks and methodologies. Among these, the Three Basic Dimensions (TBD) model, which identifies cognitive activation, classroom management, and student support as key dimensions, has gained significant recognition (see Klieme et al., 2009; Praetorius et al., 2018). While substantial evidence supports the impact of these dimensions on student outcomes, the findings are not consistently uniform across studies and contexts.

This study aims to contribute to this ongoing collaborative effort (see Charalambous et al., 2021) by examining the applicability of the TBD framework in the Czech Republic. Furthermore, we aim to explore the possibility that teaching quality does not affect all students equally. We hypothesize that its impact may differ depending on students’ prior achievement—that is, the dimensions of teaching quality may influence higher-achieving and lower-achieving students differently.

1.1 Teaching quality and its dimensions

The quality of teaching, also referred to as instructional quality, has long received substantial research attention, with numerous methods developed to study it (see, e.g., Charalambous et al., 2021). Among these methods, the Three Basic Dimensions (TBD) model has gained particular influence.

The TBD model is based on the systematizing work of Klieme et al. (2001), who extracted the three dimensions as factors when analyzing existing observational scales used in the field of teaching quality. As Charalambous and Praetorius (2018) argued, TBD is a generic framework, making it applicable in various contexts—across different school subjects, school levels, and educational systems. The TBD model identifies three crucial dimensions of teaching quality: cognitive activation, classroom management, and student support (Klieme et al., 2009). The individual dimensions are well described in a synthesizing study by Praetorius et al. (2018): (1) cognitive activation involves stimulating student thinking by building on prior knowledge, invoking higher-level thinking processes, and supporting metacognition; (2) classroom management consists of strengthening desirable student behaviors, preventing undesirable ones, and creating an environment conducive to learning without disruption and disturbance; and (3) student support involves fostering a positive climate and social interactions in the classroom, treating students with respect, and showing interest in their needs.

The TBD model implies that individual dimensions of teaching quality invoke specific learning processes in students—namely, depth of processing resulting from cognitive activation, time-on-task resulting from classroom management, and need satisfaction resulting from student support (see Alp Christ et al., 2024; Praetorius et al., 2018). Based on these constructs, it is presumed that student achievement should be positively influenced by cognitive activation and classroom management, while student support should be decisive for student motivation and interest (Alp Christ et al., 2024; Fauth et al., 2014).

There is no single instrument universally applied in studies on teaching quality; rather, researchers use a variety of measures based on different perspectives (Praetorius et al., 2017). These measures can include ratings of teaching quality by students, teachers, or external observers. Among these options, student ratings of teaching quality are most often used. They are acknowledged as the least biased, providing a long-term view on teaching and being most proximal to student learning and outcomes (see, e.g., Fauth et al., 2014; Herbert et al., 2022; Praetorius et al., 2018; Senden et al., 2023a).

The TBD has frequently served as a basis for empirical research. Most studies on this framework have been conducted in Germany (see Alp Christ et al., 2024; Atlay et al., 2019; Blume and Schmiedek, 2024; Fauth et al., 2014; Fauth et al., 2019; Fauth et al., 2021; Klieme et al., 2009; Lazarides et al., 2021; Praetorius et al., 2017; Voss et al., 2022). Furthermore, comparative studies based on international large-scale assessments have been conducted (Bellens et al., 2019; Harrison et al., 2023; Herbert et al., 2022; Liu et al., 2024; Senden et al., 2023b). However, their conclusions are rather cautious, indicating that TBD cannot be applied uniformly across countries. Therefore, there is a need for studies on non-German school systems. Although such studies are currently scarce, they are gradually emerging, with recent examples from Norway (Senden et al., 2023a; Teig and Nilsen, 2022), Croatia (Burić et al., 2023), Hong Kong (Wang et al., 2025), and Ethiopia (Sanfo and Malgoubri, 2023). This study aims to contribute to this growing body of research by providing data from the Czech Republic.

1.2 Effects of teaching quality on student achievement

In contrast to the assumption that teaching quality is essential for student learning and should be reflected in student outcomes, the empirical evidence is surprisingly inconclusive. An overview of seminal studies conducted up until 2018, as summarized by Praetorius et al. (2018), found that six out of 10 studies demonstrated a statistically significant effect of classroom management, while three out of six studies reported significant effects for cognitive activation. Although the TBD framework implies no association for student support, one out of nine studies indicated a significant effect.

Since 2019, several studies have contributed further insights with varying results: Atlay et al. (2019) in Germany reported that classroom management is positively associated with student performance, whereas cognitive activation and student support were effective only for students with high socioeconomic backgrounds. Bellens et al. (2019) in Flanders, Germany, and Norway identified classroom management as a predictor of achievement across all three educational systems, while student support did so in only one, and cognitive activation in none. Senden et al. (2023a) in Norway, investigating the link across grades and school subjects, observed that classroom management is effective across all groups, whereas the other dimensions showed mixed results. Herbert et al. (2022), analyzing data from six countries (Colombia, England, Germany, Japan, Mexico, and Shanghai), reported no universal effects of teaching quality on student outcomes. Sanfo and Malgoubri (2023) in Ethiopia noted that all three dimensions are positively associated with learning achievements. Alp Christ et al. (2024) in Germany detected no statistically significant effect of the teaching quality dimensions on student outcomes.

Disillusionment with this heterogeneity—especially regarding cognitive activation—has been expressed by numerous authors who, over the years, have conducted comparative or synthesizing studies (Bellens et al., 2019; Harrison et al., 2023; Herbert et al., 2022; Liu et al., 2024; Praetorius et al., 2018; Senden et al., 2023b). These authors call for intensifying collaborative efforts in the field and rethinking the current model with supplementary theories to advance the research (see also Alp Christ et al., 2024).

Importantly, the heterogeneity of findings is not only a measurement issue but also a theoretical one. One line of research emphasizes that instructional practices are context dependent. Building on classic work conceptualizing classroom instruction as an organizational setting (e.g., Barr et al., 1983), these authors argue that what teachers can enact is shaped by classroom composition and contextual constraints (see Bäckström, 2023; Fauth et al., 2021).

At the same time, a complementary line of work argues that the effects of teaching quality dimensions—most notably cognitively activating instruction—are conditional and depend on student-level uptake. Charalambous et al. (2025) argue that cognitive activation should be understood as an opportunity whose effects depend on students’ uptake. Drawing on Vygotsky’s (1978) notion of the zone of proximal development, this perspective implies that the same instructional move may benefit some students but not others if it falls outside their current readiness or is not taken up through active engagement.

In this paper, we integrate these perspectives but assign them different roles. We acknowledge that classroom context can shape which practices are feasible, yet our focus is on how instruction is experienced and taken up by individual students. Consistent with the idea that student-rated cognitive activation is largely student-specific (Bellens et al., 2019), and with the limited between-class variability of cognitive activation in our data, we examine whether teaching quality effects vary across groups of students with different levels of prior achievement—asking for whom cognitively activating opportunities translate into achievement gains.

1.3 Different students, different effects?

The notion that teaching quality does not affect all students equally is well-established. A growing body of research has highlighted the differential effects of instructional quality on students from various socioeconomic backgrounds (Atlay et al., 2019; Liu et al., 2024; Sanfo and Malgoubri, 2023). For instance, Atlay et al. (2019) found that students from high socioeconomic backgrounds tend to benefit more from cognitively activating instruction and a supportive classroom climate than their peers from middle- and low-SES backgrounds. Similarly, Sanfo and Malgoubri (2023) showed that classroom management and cognitive activation have stronger positive effects on students with high SES backgrounds. Drawing on international data, Liu et al. (2024) demonstrated that students from higher SES backgrounds are more likely to perceive instruction as cognitively stimulating and teachers as more supportive, while their lower SES peers tend to report less favorable experiences with instructional quality. Collectively, these studies suggest that teaching quality may unintentionally reinforce existing educational inequalities.

While socioeconomic status has received considerable attention as a moderating factor in the relationship between instructional quality and student outcomes, considerably less attention has been paid to prior academic achievement as an alternative or complementary lens. The aforementioned studies (Atlay et al., 2019; Liu et al., 2024; Sanfo and Malgoubri, 2023) focus primarily on how instructional practices may advantage or disadvantage particular social groups. However, prior achievement arguably represents a more proximal and empirically grounded indicator of educational (dis)advantage. It captures not only students’ background characteristics but also their cumulative learning trajectories and interactions with the school system. This distinction resonates with the broader peer-effects literature showing that academic composition and prior performance often account for achievement differences that are otherwise attributed to socioeconomic composition (e.g., Gutiérrez, 2023).

There are compelling reasons to expect that students with different levels of prior achievement experience and respond to instruction in qualitatively different ways. This raises important questions about whether instructional practices serve to amplify or mitigate existing disparities in achievement, and whether prior achievement might be a more sensitive lens for capturing heterogeneity in instructional effects than SES alone. At the same time, these differences unfold within group instruction: what teachers can enact is shaped not only by individual students, but also by the overall composition and needs of the class. In classes with higher average achievement, teachers may be able to sustain more cognitively demanding work and extended discussion, whereas in lower-achieving classes instruction may be more strongly constrained by prerequisite gaps and the need to secure basic engagement.

The assumption that students’ responses to teaching quality vary according to their prior achievement is supported by two strands of empirical research. First, numerous studies have shown that students’ engagement and participation in classroom activities differ systematically by achievement level (Sedova and Sedlacek, 2023; Bouton et al., 2025; Jurik et al., 2013). Second, there is evidence that teachers adapt their instructional behavior based on students’ perceived academic ability (Babad, 1993; Good et al., 2018; Snell and Lefstein, 2018). Such adaptation is therefore likely to operate both at the level of individual students and at the level of the class as a whole, potentially shaping which instructional practices are used and who benefits from them. These findings underscore the need to consider students’ prior achievement when examining the effects of teaching quality.

2 Materials and methods

The aim of this study is to explore the applicability of the TBD framework in the Czech context using 7th-grade students’ ratings of teaching quality. Moreover, we want to examine the association between TBD and student achievement in mathematics. We employ a hierarchical model with variables measured at the student level, classroom level, or both.

To structure this study, we established three research questions: RQ 1: To what extent does the three-factor model of teaching quality align with the sample of Czech 12-year-old students? RQ 2: To what extent do factors of teaching quality at both individual and classroom levels, controlling for socioeconomic status and gender, relate to student achievement in mathematics? RQ 3: Do factors of teaching quality have the same impact on student groups with different levels of mathematical ability?

This study draws on a stratified random sample of sixth/seventh grade Czech students (aged 11–12 at timepoint 1, 6th grade). The sample’s representativeness is constrained by missing data. Schools were randomly selected from different Czech regions to ensure proportional representation of the national distribution, with a target of approximately 150 schools. Within each selected school, one sixth-grade classroom was chosen at random. However, in schools offering multiple sixth-grade study programs, one classroom per program was included in the sample. Data collection was conducted via an online questionnaire administered by the Public opinion research center of the institute of sociology, Czech academy of sciences. Surveys were administered at two timepoints: between April and June 2023 (end of the 2022/23 school year) and between April and June 2024 (timepoint 2, 7th grade).

2.1 Sample

The final sample consisted of 2 609 students from 142 classrooms. Table 1 presents an overview of the sample characteristics. The gender distribution was balanced, with an equal number of boys and girls. Approximately 7% of students were of non-Czech origin, defined as having at least one parent who spoke a language other than Czech at home.

TABLE 1

Table 1. Descriptive statistics of the sample and key variables.

2.2 Measures

2.2.1 Teaching quality

We used student ratings to measure teaching quality, as previous studies have confirmed them to be effective and reliable (Fauth et al., 2014; Praetorius et al., 2017, 2018; Senden et al., 2023a). All dimensions of teaching quality were measured via student self-reports, capturing students’ perceptions of classroom processes rather than directly observed instructional practices. The questionnaire included 21 items across three subscales: classroom management (five items), cognitive activation (seven items), and student support (nine items). For cognitive activation and student support, an instrument developed by Fauth et al. (2014) was used. Classroom management was measured using PISA items. Students responded on a four-point scale ranging from 1 = strongly disagree to 4 = strongly agree (for wording, see Supplementary Appendix 1). The psychometric properties of the teaching quality model (TBD) are the focus of RQ1 and are presented in chapter 4.1.

2.2.2 Math achievement

The mathematics test utilized in this research was developed by Scio, a company that offers a system of national comparative examinations for schools in the Czech Republic. The test comprised 30 items in total, of which 17 were multiple-choice with four options each; the remaining items included open-ended, matching, completion, and ranking tasks. All items were drawn from a database used for national assessment in the respective academic year (aligned with the curriculum for Grades 6 and 7). The objective of the test was to effectively discriminate between different levels of mathematical proficiency. Consequently, the test incorporated a range of thematic items (such as arithmetic, geometry, and application tasks), varying levels of difficulty (based on psychometric properties from national testing), and diverse objectives according to Bloom’s taxonomy (ranging from memorization to comprehension to application). Students’ total test scores in mathematics were estimated using Item Response Theory (IRT) models. To estimate IRT scores, we applied a mixed-model approach, using a three-parameter logistic (3PL) model for multiple-choice items and a two-parameter logistic (2PL) model for other item types (See Supplementary Appendix 3). Model fit was evaluated using conventional indices and indicated a strong fit [M²(388) = 1006.533, p < 0.001; RMSEA = 0.025; SRMSR = 0.025; TLI = 0.987; CFI = 0.988]. These values confirm that the mixed IRT model fits the data well and supports the validity of the estimated ability scores. Examples of the tasks and their layout can be found in Supplementary Appendix 2. Table 2 summarizes the key parameters of the IRT score models. Anchor items in mathematics cover the middle range of difficulty and show good discrimination. Although some multiple-choice items in the z tests have c-parameter estimates close to zero (Table 2), the proposed mixed model fits the data significantly better than the more parsimonious two-parameter model, which fixes the guessability parameter at zero for all items [Δχ²(34) = 596.01, p < 0.001]. Additionally, the information criteria values (AIC and BIC) are lower for the mixed model than for the two-parameter model.

TABLE 2

Table 2. IRT parameter estimates for items in the mathematics tests.

2.2.3 Control variable: socioeconomic status

Index of the students was constructed using the principal component analysis (PCA) method, specifically as the first principal component derived from indicators of parental education, parental occupation, household assets, and the number of books in the household. This approach is similar to that used in the Programme for International Student Assessment (PISA) survey (Avvisati, 2020), where the number of books is also considered, alongside other material and cultural assets, as part of the wealth indicator. Table 3 reports the factor loadings of each SES indicator and the proportion of variance in the indicators explained by a common factor. The proportion of variance explained is 48%.

TABLE 3

Table 3. Standardized factor loadings of indicators of SES.

2.2.4 Missing

The proportion of missing data across all measures was in the usual range, with no indication of systematic missingness. To account for these missing values, we applied full information maximum likelihood (FIML), ensuring that all available data contributed to the analysis.

2.3 Analytical procedure

To investigate the factorial structure (RQ1), three distinct confirmatory factor analyses were conducted. The overall goodness-of-fit was assessed through absolute fit indices, adhering to the conventional cutoff values established by Hu and Bentler (1999). Specifically, the standardized root mean square residual (SRMR) was considered acceptable with values close to or below 0.08, the root mean square error of approximation (RMSEA) with values near or below 0.06, and both the comparative fit index (CFI) and the Tucker-Lewis index (TLI) with values close to or exceeding 0.95.

The second and third research questions were examined using a multilevel structural model designed to assess the effects of teaching quality at both individual and classroom levels on grade 7 mathematics achievement. The model controlled for students’ socioeconomic status (SES) and gender, with prior mathematics achievement in grade 6 included as a key covariate to account for existing knowledge and isolate the net effect of teaching quality. Model fit for this analysis was also evaluated using the established criteria from Hu and Bentler (1999).

All data management was conducted using IBM SPSS Statistics Version 30, and the analyses were performed with the lavaan package (Rosseel, 2012) in the R environment.

2.4 Procedure and ethics

Data collection took place during classroom-wide assessments conducted by trained staff following standardized instructions. Students received guidance on how to approach the test items. To accommodate language and reading difficulties, the items were read aloud to the entire class, and students were given time to respond after each item.

Participation in the research required obtaining informed consent from the students’ legal guardians. The research was approved by the Ethics Committee of the Institute of Sociology of the Czech Academy of Sciences, which oversaw the data collection.

3 Results

3.1 Does the three-factor model of teaching quality align with the data?

Student ratings of teaching quality offer two distinct perspectives. At the individual level, they capture students’ personal perceptions of teaching quality, whereas, when aggregated at the classroom level, the average rating represents a collective perception of teaching quality. In previous research, the factor structure of the teaching quality model has been demonstrated across various samples. We followed the same procedure to assess whether the three-factor structure at the student and class levels aligned with data from Czech schools. However, as shown in Table 4, this structure did not provide a good fit. Thus, the three-factor model does not adequately reflect students’ responses. The cognitive activation factor was particularly problematic, as the items failed to capture a single latent construct (see Supplementary Appendix 1). Based on the wording of the items and an exploratory factor analysis, we hypothesized two distinct dimensions of cognitive activation: cognitive activation through learning tasks (CAT) and cognitive activation through interaction with the teacher (CAI). We incorporated both emergent cognitive activation factors into a model of teaching quality. The fit of the four-factor model is reported in Table 4.

TABLE 4

Table 4. Results of the model comparison between the three-factor and four-factor TBD models.

A comparison of the fit indices for the two models clearly indicates that the four-factor model of teaching quality at both the individual and classroom levels provides a significantly better fit to the data. The standardized factor loadings at both levels are sufficiently high (ranging from 0.365 to 0.674) and statistically significant (p < 0.001), suggesting that the items effectively capture the latent factors. For each instructional quality factor, we calculated intraclass correlation coefficients: ICC 1, representing the proportion of total variance explained by the cluster structure, and ICC 2, indicating the reliability of the aggregate variables. The analysis of ICC 1 suggests that the factors of teaching quality exhibit varying levels of between-class variability. The CAT and CAI factors have very low ICC 1 values (0.05 and 0.10, respectively), indicating minimal differences between classes. This suggests that these factors function more as individual characteristics, making their aggregation less meaningful. A similar pattern has been observed in other educational contexts. For example, Senden et al. (2023a), using a Norwegian sample, reported a relatively low ICC for cognitive activation (0.10), compared to higher ICCs for classroom management (0.25) and supportive climate (0.14). This supports our interpretation that cognitively activating instruction tends to vary more within classrooms than between them. In contrast, classroom management and supportive climate demonstrate satisfactory ICC 1 values (0.27 and 0.18, respectively). Moreover, the ICC 2 values indicate that these latent variables at the class level are highly reliable (0.98 and 0.96, respectively). Therefore, in the subsequent analysis, we consider CAT and CAI only at the individual level, while classroom management and supportive climate are analyzed at both levels.

3.2 To what extent do factors of teaching quality relate to student achievement in mathematics?

After verifying the factorial structure of teaching quality, we sequentially constructed four structural models to examine predictors of mathematics achievement in grade 7. The results are presented in Table 5.

TABLE 5

Table 5. Results for the structural equation model relating the factors of teaching quality to achievement in math.

The first two models (1a and 1b) serve as control models, including only background variables. Model 1a tests the effects of student demographics (gender and socioeconomic status), explaining 11% of the variance in achievement. Both SES and gender were significant predictors, with girls scoring slightly lower and SES showing a positive association with performance. Model 1b adds prior mathematics achievement (Math1), which substantially increases the explained variance to 46%. As expected, prior achievement is the strongest predictor (β = 0.651, p < 0.001), followed by SES (β = 0.101, p < 0.001), while gender has a weaker effect (β = –0.067, p < 0.05). Including teaching quality factors in the subsequent models further increases the explained variance to 48% in Model 2 (individual-level teaching quality) and 49% in Model 3 (with additional classroom-level teaching quality factors).¹

At the individual level, classroom management, supportive climate, and cognitive activation through learning tasks (CAT) are not statistically significant predictors. Only cognitive activation through interaction with the teacher (CAI) demonstrates a significant positive effect (β = 0.101, p < 0.05). At the classroom level, supportive climate (SC_class) is not significant, whereas classroom management (CM_class) shows a modest but significant positive effect (β = 0.091, p < 0.001). In the final model (Model 3), non-significant predictors were removed, resulting in a model that fits the data well (CFI = 0.982; TLI = 0.973; RMSEA = 0.040; SRMR = 0.0031).

3.3 Do factors of teaching quality have the same impact on student groups with different levels of mathematical ability?

We use the significant teaching quality factors to test the final assumption that their effects may vary based on students’ prior mathematical achievement. Based on their prior achievement (Grade 6), we divided students into quartiles and constructed a separate structural model for each group (see the final model from RQ2). Students were assigned to quartiles using the full distribution of individual prior achievement scores, while maintaining their original classroom assignments. This means that class membership remained intact, and the multilevel structure of the data was preserved. We chose quartile-based grouping to balance interpretability and statistical power, though alternative approaches—such as median splits or continuous moderation—were considered. The results in the form of standardized regression coefficients are presented in Table 6.

TABLE 6

Table 6. Results of comparisons of structural models examining the effect of instructional quality on mathematics achievement across four quartiles of mathematical skills.

The results indicate that teaching quality factors operate differently across student groups. For the students with the lowest prior mathematics achievement (Quartile 1), none of the factors are significant. The same applies to students with slightly below-average results (Quartile 2). However, for students with slightly above-average results (Quartile 3), CM at the whole-class level has a significant positive effect. In contrast, for the highest-achieving students (Quartile 4), CM_class is no longer significant, while CAI has a positive effect.

4 Discussion

In this study, we examined the extent to which the TBD model of teaching quality aligns with data collected from Czech 7th grade students. Our findings suggest that the model fits the data well, but in a slightly revised form—more accurately described as a “four basic dimensions” model. Specifically, the cognitive activation factor splits into two distinct dimensions: cognitive activation through learning tasks (CAT) and cognitive activation through interaction with the teacher (CAI). A revision of the original three-factor structure of the TBD framework had already been proposed by Bellens et al. (2019), who argued that further differentiation within the cognitive activation (CA) scale was necessary in order to establish a valid and reliable construct. In line with our findings, they suggested that the CA factor could be understood as consisting of two components: one capturing support for learning and the other reflecting the triggering of higher-order thinking processes in students. Our analysis supports and extends this claim, providing additional empirical evidence for the distinctiveness of these two components.

This distinction also invites a theoretical interpretation. Cognitive activation—understood as stimulating student thinking by building on prior knowledge, eliciting higher-order thinking processes, and supporting metacognition (see Praetorius et al., 2018)—appears to unfold through two qualitatively different pathways. The interplay of CAT and CAI can be meaningfully interpreted through the lens of sociocultural theories of learning (Vygotsky, 1978), which emphasize that effective learning occurs not solely through individual cognitive effort but through social interaction and guided participation.

In this framework, cognitive activation through learning tasks (CAT) can be viewed as a form of challenge. It refers to the provision of complex, thought-provoking tasks that require students to stretch their current understanding, engage in abstract reasoning, or solve non-routine problems. These tasks push students to operate at the edge of their competence, which is essential for growth—but only if appropriate support is also available. In a classroom, students differ in prior knowledge, implying multiple zones of proximal development that instruction must address under group instruction. This heterogeneity can steer and limit what teachers can enact, so the same cognitively demanding task may be optimally challenging for some students but out of reach for others.

On the other hand, cognitive activation through interaction with the teacher (CAI) represents a form of support. It encompasses teacher practices such as asking questions, probing students’ reasoning, and engaging in instructional dialogue. These interactions serve to scaffold students’ thinking, help them articulate and refine their ideas, and maintain their engagement when grappling with cognitively demanding material. From an opportunity–use perspective (Charalambous et al., 2025), such interaction can be seen as a mechanism that supports students’ uptake of cognitive challenges—helping them actually use the opportunity rather than disengage from it.

Taken together, CAT and CAI reflect the two key elements Vygotsky (1978) described as necessary for learning within the zone of proximal development: intellectual challenge and responsive support from a more knowledgeable other. In this sense, the differentiation of the cognitive activation construct in our model not only improves its empirical validity but also deepens its theoretical significance.

In this study, we further investigated the extent to which dimensions of teaching quality at both the individual and classroom levels are related to students’ mathematics achievement. Structural equation modeling revealed that only some aspects of teaching quality—namely, classroom management and student support—operate meaningfully at the classroom level. In contrast, both types of cognitive activation appear to be primarily individual-level phenomena: their variance is only marginally explained by classroom membership. This suggests that students within the same classroom perceive the difficulty of assigned tasks (CAT) and the degree of stimulation through teacher interaction (CAI) in markedly different ways. This pattern fits the idea that instruction is enacted at the classroom level under group-instruction constraints (Bäckström, 2023; Fauth et al., 2021), while students differ in how they experience and take up the same learning opportunities (Alp Christ et al., 2022; Charalambous et al., 2025). In this sense, classroom context shapes what is feasible to enact, but cognitive activation—as captured by student ratings—primarily reflects student-level differentiation within that shared setting.

Previous studies conducted in other contexts support the idea of such heterogeneity in perceived cognitive challenge and teacher support (see, e.g., Ben Maad, 2012; Tavakoli, 2009), and demonstrate that students’ perception of task difficulty plays a crucial role in how systematically and persistently they engage with the task (Cornelisz and van Klaveren, 2018; Nawaz et al., 2021).

For these reasons, we modeled classroom management and student support at the classroom level, whereas CAT and CAI were treated as individual-level variables. Our analysis showed that two factors significantly predicted student achievement: classroom management and cognitive activation through interaction with the teacher. Students perform better when they are in classrooms with better management and when they experience more individualized interaction with the teacher that stimulates their thinking.

These findings are broadly in line with the theoretical assumptions of the TBD model, which posits that student achievement is positively influenced by cognitive activation and classroom management, while student support primarily fosters motivation and interest (Alp Christ et al., 2024; Fauth et al., 2014; Praetorius et al., 2018). However, our results should not be interpreted as a straightforward confirmation of these assumptions, for two main reasons.

First, the two cognitive activation factors operate in very different ways. In this respect, our findings are consistent with those of Bellens et al. (2019) and Liu et al. (2024). According to our data, it is only CAI that has a positive effect. The effect of CAT—although statistically insignificant—appears to be negative. While the TBD model assumes that higher cognitive activation is beneficial for student achievement, our findings suggest that when students perceive tasks as highly difficult, their performance does not improve. Studies by Cornelisz and van Klaveren (2018) and Nawaz et al. (2021) help us interpret this result: students tend to be most productive when they perceive tasks not as very difficult but as moderately challenging. Thus, when it comes to cognitive activation, more is not necessarily better. This aligns with Charalambous et al. (2025), who argue that cognitive activation should not be expected to show robust direct effects unless the opportunities are adapted to students’ readiness and actually used. From this perspective, highly demanding tasks may be ineffective—or even detrimental—when they exceed students’ current readiness and are therefore not taken up through deep processing. It appears more effective to consider the optimal level of cognitive activation through learning tasks, while also maintaining cognitively rich interaction with the teacher.

Second, as shown in our follow-up analyses, neither classroom management nor cognitive activation through teacher interaction work universally for all students. For students with low prior achievement, neither factor had any significant impact—regardless of the classroom context. For slightly above-average students, classroom management proved beneficial—quiet and focused classroom environments helped them perform better. Related work highlights the importance of orderly, low-disruption learning environments for disadvantaged students (Yang Hansen et al., 2025), although differences in measures and levels of analysis may explain why we do not observe such effects for students in the lower prior-achievement quartiles. For high-achieving students, CAI made the difference: cognitive stimulation supported their learning, not in the sense of providing more challenge, but as support—in line with Vygotsky’s (1978) notion of scaffolding within the zone of proximal development. In the terms of Charalambous et al. (2025), CAI may be particularly consequential when it helps students use cognitively demanding opportunities productively; for lower-achieving students, the same opportunities may remain outside their current readiness or may not be taken up, which would attenuate observable effects.

Taken together, our findings lend further support to concerns raised in previous research (Atlay et al., 2019; Sanfo and Malgoubri, 2023) that teaching quality, while generally seen as a key to improving learning outcomes, may in fact contribute to widening educational inequalities. In our study, the relationship between teaching quality and student achievement was positive only for students who were already above average in prior achievement. At the same time, we observed a particularly strong direct effect of prior achievement on current achievement. This suggests that students with lower initial performance face a double disadvantage: they not only start from a lower academic baseline, but also benefit less—or not at all—from the available teaching quality.

These findings raise a critical point: teaching quality is not a universally effective mechanism. Although classroom management and teacher-student interaction (CAI) contribute positively to student achievement, their impact is limited to students who already demonstrate a certain level of prior success. Students with lower prior achievement—who arguably have the greatest need for instructional support—appear to benefit neither from structured classroom environments nor from cognitively stimulating teacher interactions. At the same time, cognitive activation through tasks (CAT), while intended as a form of instructional challenge, may even hinder learning if the task is perceived as too difficult. These patterns highlight an important limitation: current teaching practices, even when generally high in quality, do not automatically address the needs of all students.

This observation aligns with broader research showing that education systems often reproduce, rather than reduce, inequality—especially in contexts like the Czech Republic, where the influence of socioeconomic background on student outcomes remains particularly strong (Bodovski and Byun, 2017). Students’ educational development is shaped by a process of cumulative (dis)advantage (DiPrete and Eirich, 2006), where early setbacks—academic or social—can determine long-term trajectories. If teaching quality fails to reach those who need it most, it risks becoming a mechanism that reinforces rather than remedies educational gaps (Carter et al., 2020).

In sum, teaching quality matters—but it matters differently. Our study underscores the need for more differentiated, responsive teaching practices, particularly in systems where student diversity is high and structural inequalities persist. Future research should focus on how teaching can be adapted and individualized, especially to support low-achieving students and avoid reinforcing existing disparities. Only then can the full potential of teaching quality as an equalizing force in education be realized.

4.1 Implications

The findings of this study carry several important implications for educational practice, particularly in the areas of teacher professional development and instructional design. While the concept of teaching quality remains central to improving student achievement, our results show that its effects are not equally distributed across learners. This pattern suggests that traditional approaches to improving instructional quality—however effective in general—may not suffice to reduce educational inequality (Fauth et al., 2019; Blume and Schmiedek, 2024).

This underscores the need to shift from a one-size-fits-all approach to a differentiated and responsive model of teaching. Teachers must be supported not only in delivering high-quality instruction, but also in learning how to adapt it to the diverse needs and learning trajectories of their students. Supporting students who benefit less under current instructional conditions requires an intentional combination of challenge and scaffolding, as described in sociocultural theories of learning. Our study indicates that it is not cognitive challenge per se that promotes achievement, but rather the way it is mediated and supported through teacher-student interaction. For teacher training, this means placing stronger emphasis on interactional repertoires to ensure that cognitive activation is accessible rather than selectively beneficial.

Addressing this challenge calls for reimagining teacher professional development. Teachers should be offered opportunities to build expertise in recognizing and responding to student variability, particularly in how students perceive and experience the same instructional environment. Recent research has shown that professional development interventions, when grounded in a coherent pedagogical framework, can produce significant improvements in teaching quality (Gore et al., 2017, 2021). To be effective, professional learning must go beyond generic instructional strategies and engage teachers in structured reflection, peer observation, and evidence-based dialogue about student thinking and learning. Practically, professional learning activities can be organized around identifying which students benefit least from current classroom interaction patterns and rehearsing targeted adjustments that make high-demand tasks workable for a wider range of learners. Investing in such models of professional learning—especially in systems where structural inequalities are pronounced—can help ensure that teaching quality becomes a more equitable resource, reaching those who need it most.

4.2 Limitations

The findings of this study are specific to the Czech educational context and must be interpreted with this in mind. Czech classrooms are shaped by particular interactional norms and relatively traditional teaching practices, and many teachers may not yet be fully equipped to address student diversity through differentiated instruction (Obrovská et al., 2024). This may partly explain why lower-achieving students in our study did not benefit from teaching quality in the same way as their higher-achieving peers.

It remains an open question whether similar patterns would be observed in educational systems that operate under different cultural, pedagogical, and institutional conditions. Research from other contexts suggests that instructional constraints and opportunity-to-learn conditions may differ substantially across systems, potentially leading to different patterns of inequality (e.g., Bäckström, 2023). As noted by Herbert et al. (2022), cross-national variation in findings may be influenced by cultural differences in instructional practices and teacher-student interaction norms. Therefore, further research is needed to examine how the relationship between teaching quality, student participation, and achievement unfolds in other educational contexts, particularly in systems where teachers are more accustomed to differentiated instruction and inclusive pedagogies.

6 Conclusion

This study contributes to the growing body of international research examining the structure and impact of teaching quality by applying the Three Basic Dimensions (TBD) framework to the Czech educational context. Our findings support a revised, four-dimensional model in which cognitive activation separates into two distinct components: task-based challenge (CAT) and teacher-mediated support (CAI). This differentiation not only improves the empirical validity of the model but also deepens its theoretical alignment with sociocultural learning theories.

The results show that teaching quality does relate to student achievement—but not equally for all students. While classroom management and cognitively stimulating teacher interaction (CAI) were positively associated with student performance, these benefits were largely limited to students with higher prior achievement. For lower-achieving students, even high-quality instruction appeared to have limited or no effect, and cognitively demanding tasks could be counterproductive when perceived as too difficult.

These findings suggest that teaching quality is not inherently equitable and may even reinforce existing disparities if instruction is not tailored to students’ prior learning trajectories. They highlight the urgent need for professional development that equips teachers to implement differentiated and responsive teaching practices. Only by addressing the variability in how students perceive and benefit from instruction can we unlock the full potential of teaching quality as a driver of educational equity. At the same time, these associations are likely to be shaped by contextual conditions (e.g., curriculum, classroom norms, and school-level supports), and therefore the strength and pattern of results may differ across settings.

Data availability statement

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found at: https://archivdv. soc.cas.cz/dataset.xhtml?persistentId=doi:10.14473/CSDA/I17GF8; https://archivdv.soc.cas.cz/dataset.xhtml?persistentId=doi:10.14473/CSDA/O9POT7.

Ethics statement

The studies involving humans were approved by the Ethics Committee of the Institute of Sociology of the Czech Academy of Sciences. The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent for participation in this study was provided by the participants’ legal guardians/next of kin.

Author contributions

KS: Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Writing – original draft, Writing – review & editing. MS: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review & editing.

Funding

The author(s) declared that financial support was received for this work and/or its publication. This work was supported by the NPO “Systemic Risk Institute” number LX22NPO5101, funded by European Union—Next Generation EU.

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. Generative AI tools (ChatGPT by OpenAI) were used solely for language editing and improving the clarity and fluency of the manuscript. The authors reviewed and approved all AI-assisted edits and take full responsibility for the content and conclusions of the article.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/feduc.2025.1712000/full#supplementary-material

Footnotes

1. ^The total explained variance (R²) for mathematics achievement in grade 7 was 0.491. This reflects the combined explanatory power of the teaching quality factors, prior achievement, and control variables at both the individual and classroom levels.

References

Alp Christ, A., Capon-Sieber, V., Grob, U., and Praetorius, A.-K. (2022). Learning processes and their mediating role between teaching quality and student achievement: A systematic review. Stud. Educ. Eval. 75:101209. doi: 10.1016/j.stueduc.2022.101209