Assessing Teacher–Student Interactions in Physical Education for Sustainable Development: Validation of the CLASS-S Tool in the Polish Educational Context

Abstract

Ensuring healthy living and promoting well-being (Sustainable Development Goals 3) is a key goal of school Physical Education. As many students in Poland perceive these lessons as unengaging and ineffective, there is a need to explore new educational strategies. One such approach is teaching through interaction, which emphasises the quality of teacher–student relationships. This study aimed to adapt and validate the Classroom Assessment Scoring System—Secondary for use in Polish Physical Education settings, considering the cultural context of Polish schools and the challenges of applying interaction-based frameworks in a traditionally structured education system. Fifty teachers from 24 schools recorded two lessons each, resulting in 100 video recordings. Lessons were analyzed using a structured observational protocol. Teacher–student interactions were recorded and coded by two trained observers. Confirmatory factor analysis confirmed excellent model fit (χ²/df = 1.158; CFI = 0.996; TLI = 0.998), and internal consistency was supported by Cronbach’s alpha. Inter-rater reliability was high, with intraclass correlation coefficients ranging from 0.928 to 0.946. The modified tool is both valid and reliable. Its application can enhance the evaluation and professional development of Physical Education teachers and support the integration of health- and education-related goals in school practice and contribute to achieving SDGs 3 and 4.

1. Introduction

The 2030 Agenda [1] emphasises the importance of education, establishing it both as a stand-alone goal (Sustainable Development Goal 4) and as a tool for achieving the other Sustainable Development Goals (SDGs). As an integral component and catalyst for transformation, education provides the foundation for achieving a sustainable future. Research indicates that education for sustainable development is poorly implemented in Physical Education (PE) lessons. The curricula for this subject contain few references to this concept, and the teachers themselves, although sometimes willing to implement it, have no influence on the content of these curricula [2,3,4]. In Poland, the teaching standards do not mention the SDGs at all, so consequently, no one requires their implementation [5,6]. Bridging these shortcomings is attempted by arguing why it is worth implementing in PE programmes and how to do it [2,3]. Regardless of any conscious or unconscious reference to the SDGs, ensuring healthy living and promoting well-being for all (SDG 3) [7] should be one of the main goals of institutionalised school PE. The links between good quality PE and the SDGs, especially SDG 3: Good health and quality of life, and SDG 4: Good quality education [8], are obvious. School Physical Education significantly improves health and develops healthy lifestyles, which is key to reducing the risk of lifestyle diseases, including obesity, heart disease, or diabetes. If SDG 4 is about eliminating barriers to education and promoting equal access to learning, then PE, if well planned and delivered, can become an inclusive space for all students, regardless of their fitness level, gender, social background, or degree of disability.

The 2018 Report on the Implementation of the Sustainable Development Goals in Poland [9] states that measures of the health status of the population indicate that the Polish population is below the average for the EU20 as a whole. Hence, one of the priorities under SDG 3 was to reduce the incidence of and mortality from civilisation diseases and to increase health awareness, improve nutrition, and increase physical activity among the population. In turn, the defined priorities for SDG 4 were to improve the innovation of education through reforming the education system. Ensuring the health of citizens took on particular importance after the outbreak of the COVID-19 pandemic. At that time, a number of negative health trends related to the population’s lifestyle, including low physical activity, were diagnosed. In the next Report on the implementation of the 2023 Sustainable Development Goals in Poland [10], prevention and health promotion, especially in the fight against being overweight and obesity, became a priority again. The way to achieve this goal was through increased physical activity of the population. The challenge for SDG 4 was identified as the development of inclusive education and raising the quality of education through the development of educational innovation.

However, if students in Poland report that PE lessons are boring and unattractive, and thus ineffective, it seems important to look for new educational strategies. The Sustainable Development Goals (SDGs 3 and 4) can be achieved by implementing the ‘teaching through interaction’ (TTI) concept into the PE process [11]. The Classroom Assessment Scoring System (CLASS) tool [12] constructed based on the TTI operationalises teacher–student interactions in three main domains: (1) Emotional support; (2) Class organisation; and (3) Instructional support. The domain of emotional support fits with the goals of SDG3, as it emphasises the importance of building a positive climate in the classroom, which can contribute to reducing student stress, anxiety, and burnout, thus building mental health. Building good social relationships (teacher–student, student–student) by recognising and responding to students’ social, health, behavioural, and cognitive needs also translates into student well-being. The TTI model assumes that the quality of education does not depend solely on the curriculum delivered and the methods used, including modern technology and teaching resources, but is the result of quality interaction between teacher and student. Emotional and instructional support of the student as well as good organisation of the educational environment fit in with SDG 4. Teacher–student interactions can promote more inclusive education, for example, by tailoring approaches to students from different social backgrounds. TTI promotes an approach that minimises educational barriers, which is key to achieving SDG 4. In addition, instructional support helps students develop key 21st-century competencies such as critical thinking, collaboration, creativity, and higher-order thinking, which is relevant to the Global Goals for Education. The TTI model supports both the UN Sustainable Development Goals on health (SDG 3) and education (SDG 4) by emphasising the importance of human relationships in the learning process. Better quality classroom interactions lead to better student well-being and more effective education, which is in line with the UN’s global priorities of equality, well-being, and quality education.

Despite this relevance, the notion of teaching through interaction is relatively unexamined with regard to physical education teacher education (PETE). There remains a consistent void in the literature about how prospective PE teachers are prepared to develop quality interactions with students and how tools such as CLASS can facilitate their professional development. Such a deficit could be compensated for by the inclusion of interaction-focused methods in teacher education programmes to ease the way to initial teacher training or in-service programmes. Significantly, to date, no research has utilised the CLASS model to investigate the quality of teacher–student interactions during activities in PE lessons. The CLASS tool has been developed and utilised mainly in theoretical classroom settings. This research thus contributes to the present research base by investigating whether teacher–student interactions in movement-based PE classes are akin to interactions seen in classical academic environments. In doing so, it not only contributes towards the tool’s validity in a new educational context but also sheds light on the applicability of the TTI framework in general. The findings will consequently have implications for not only the preparation of Physical Education instructors but also for the further evolution of the CLASS tool itself as well as for the development of interaction-based pedagogical theory.

The application of this concept in Poland requires the preparation of appropriate diagnostic tools. While different versions of CLASS have been used in many countries and validated in different cultural settings, CLASS-S has rarely been used outside the US. We, therefore, have few studies confirming its theoretical relevance in settings other than the US education system [13,14,15,16]. In Poland, this tool in the CLASS-UE version [17] was used once in the Study of School Conditions of Educational Effectiveness (SUEK) conducted in elementary schools from 2010 to 2014 under the direction of Dolata [18], but we do not have validation of this tool or detailed information about the course of the study and the results obtained. The CLASS-S version has not been used by anyone in Poland. Early studies using CLASS-S indicated conflicting findings regarding the number of domains and the assignment of dimensions to them. Allen et al. [19] confirmed the superiority of the three-factor structure, which was later confirmed in a large group of middle and high school classes in the US by Hafen et al. [20], Virtanen et al. [14]; Razo and Peralta [15]; and Westergård et al. [16]. It should be noted, however, that not all studies have confirmed this hypothetical structure, and the number of dimensions within domains has varied under different cultural conditions. Malberg et al. [13] found that the three-factor structure was a poor fit to the data. In contrast, a study by Kane et al. [21] found a single-factor solution to be the best fit. A review of the literature did not reveal that the CLASS-S tool has been used to date in research on the quality of teacher–student interactions during PE lessons. Research to date has mainly focused on analysing interactions in the context of theoretical subjects. The lack of reports on the use of the CLASS-S tool during sports lessons means this study will be an important contribution to research on the use of the tool in different educational settings and environments. Furthermore, it is worth examining whether the cultural characteristics of Polish education—such as its traditionally teacher-centred approach and limited emphasis on student voice—may influence how teacher–student interactions are observed and interpreted using internationally developed tools like CLASS-S. Hence, this study aimed to validate the Classroom Assessment Scoring System—Secondary (CLASS-S) for use in Polish Physical Education lessons. A specific research question was whether the original three-factor structure of the CLASS-S would be confirmed in the Polish educational context and within the domain of Physical Education.

2. Materials and Methods

2.1. Participants

Participants in this study were recruited from three Polish voivodeships: Mazowieckie, Opolskie, and Małopolskie. A purposive sampling strategy was applied to ensure representation from diverse educational regions. Schools were nominated for the study by the authorities in charge of their schools (county heads, city presidents, or district mayors). The goal was to increase the contextual diversity of the sample while maintaining feasibility in data collection. School principals then recruited teachers from among the teaching staff. Teachers, adult students, and parents of underage students gave their written consent to participate in the study. All participants were informed about the purpose of the study, how and under what conditions personal data would be processed, and the possibility of withdrawing consent at any stage of the study. The study also received ethics committee approval. Fifty teachers (24 women and 26 men) from 24 secondary schools participated in the study. The mean length of service of the teachers was 17.5 years (SD = 9.95). A portion of the students, specifically 23% (n = 169), did not provide consent to take part in the study. Students without written consent did not participate in the recorded lessons and were supervised by other teachers during this time. Ultimately, 560 students participated in the study, 318 boys and 242 girls. Numbers in the recorded lessons ranged from 4 to 22 pupils (M = 11.65, SD = 3.95). Teachers were free to choose the forms of physical activity, objectives, content, methods, and teaching resources. Detailed information on the characteristics of the subjects and the conditions of the study is presented in

Table 1. Participant demographics and conditions of observed lessons.

		Number (%)
Gender of teachers	Men	26 (52%)
	Women	24 (48%)
Teacher’s professional promotion grade	beginner	8 (16%)
	appointed	7 (14%)
	qualified	35 (70%)
Gender of students	male	318 (56.8%)
	female	242 (43.2%)
Form of physical activity during observed lesson	Team games	62 (62%)
	Gymnastics/Dance/Fitness	13 (13%)
	Ex. general development/gym	19 (19%)
	Individual sports (LA, swimming)	6 (6%)
Gender of students during observed lessons	male	40 (40%)
	female	32 (32%)
	mixed-gender groups of boys and girls	28 (28%)

.

2.2. Measurement

The Classroom Assessment Scoring System—Secondary tool was used to assess teacher–student interaction [12]. The criterion for selecting a version of the CLASS tool was the age of the students. CLASS-S assesses the quality of teacher–student interaction in three domains: emotional support, organisational support, and instructional support—which consist of 11 more specific dimensions.

The emotional support domain contains dimensions relating to the teacher’s ability to create and maintain a supportive social environment that fosters engagement in learning. These activities can be enhanced through interactions that foster student autonomy, the building of good relationships, and a sense of competence. In the CLASS-S protocol, the emotional support domain consists of three dimensions: positive climate, teacher sensitivity, and consideration of the student’s perspective.

The classroom organisation domain refers to the teacher’s ability to organise students’ work. It includes good management of problematic/disruptive behaviour and the extent to which behavioural rules and procedures are followed. This domain pays attention to time management, maximising learning time and minimising unnecessary interruptions. This domain consists of three dimensions: behaviour management, productivity, and negative climate.

The third domain assesses the teacher’s ability to support learning. While emotional support assesses the social climate in the classroom, learning support assesses the quality of the teacher’s actions in the use of methods and techniques for teaching dialogue and feedback, and the way the content is organised. This domain includes the dimensions of teaching methods and forms, understanding the content, quality of feedback, analysing and solving problems, and teaching dialogue.

The Student Engagement dimension, while part of the CLASS-S framework, does not belong to any of the three core domains. Instead, it captures the overall level of student participation and involvement in classroom activities. Consistent with prior validation studies [14,19,20], this dimension was excluded from the current analysis.

Each dimension within the CLASS-S system comprises a set of indicators that characterise effective teacher–student interactions. These indicators are operationalised through clearly defined and observable behavioural markers, which guide trained observers in evaluating classroom dynamics. The three CLASS-S domains (emotional support, classroom organisation, and instructional support) are each composed of three to five dimensions. Observers assign scores to each dimension on a 7-point scale, where 1–2 indicates low quality, 3–5 moderate quality, and 6–7 high quality of interaction. Detailed descriptions of behavioural markers, scoring guidelines, and the process for aggregating ratings from indicators to dimension-level scores are available in the manual Classroom Assessment Scoring System—Secondary [12].

2.3. Procedure

Research data were collected from January 2023 to March 2024 during 50 classes in secondary schools. Two PE lessons for each teacher with one class were recorded over one teaching week, so a maximum of 5 days apart. The lessons were recorded by a research assistant with the camera positioned so that it captured the teacher and all practitioners while not interfering with the lessons; once the camera was set up, the research assistant would leave the classroom area. The teacher was equipped with a RODE Wireless GOII microphone to record verbal messages. A total of 100 lessons were recorded, lasting an average of 37:17 min (min. 32:11, max. 42:23 min). Each lesson was divided into two observation cycles according to the CLASS-S protocol [12]. The first cycle was from 00:00 to 15:00 min of recording, and the second cycle from 15:00 min to 30:00 min, as there is evidence that the first 30 min of a lesson is representative of the quality of the whole lesson [16,22]. Two certified observers independently assessed all lessons from the video recordings. The observers attended the CLASS 2008 Observation Training Secondary course delivered by Teachstone (Charlottesville, WV, USA) and subsequently passed the certification exams (CLASS 2008 Observer Secondary). To ensure calibration prior to formal coding, the observers rated four pilot video lessons not included in the study sample. Each teacher was recorded twice and each lesson was divided into two observation cycles wherein each observer coded 200 observation sheets (a total of 400 observation cycles were assessed). This means that each teacher was assessed 8 times by two observers. For each teacher, the score for each CLASS-S dimension was calculated by first averaging the ratings from four observation cycles (2 cycles per lesson × 2 lessons) provided by a single observer. Then, the dimension scores from both observers were averaged to obtain a final value per teacher. Domain-level scores (emotional support, classroom organisation, instructional support) were computed as the mean of the scores from the dimensions assigned to each respective domain.

2.4. Analysis Strategy

An analysis of the relevance and reliability of the CLASS-S tool under Polish [education] conditions and for the specific needs of PE was performed in three steps. In the first step, a confirmatory factor analysis (CFA) was conducted to confirm the structure of the original tool, using the maximum likelihood (ML) method. Five indicators were used as a criterion for goodness of fit: χ², standardised root mean square of residuals (SRMR), root mean square of approximation error (RMSEA), comparative fit index (CFI), and Tucker–Lewis index (TLI). The cut-off values for well-fitted models were as follows: χ² = ns (p > 0.05), SRMR < 0.08, RMSEA close to 0.06, CFI and TLI close to 0.95 and TLI [23]. Different approaches to the CLASS-S factor structure were analysed. First, the fit of the collected data to the original three-factor structure was verified [12]. A one-factor structure was also considered as well as a two-factor structure combining emotional support and instructional support in one factor, leaving classroom organisation as a separate factor. Finally, the fit of the modified three-factor structure was checked based on the observations of Polish PE teachers.

In the second step of the CLASS-S validation, the reliability of the scales was examined using Cronbach’s alpha index, composite reliability (CR), and average variance extracted (AVE). Its reliability was then examined by analysing the consistency of the two observers’ ratings. For this purpose, an intraclass correlation coefficient (ICC) and 95% confidence interval (95% CI) were used. A 2-factor random-effects model and absolute concordance were used. The score was interpreted according to the more restrictive criteria of Koo and Li [24]: less than 0.5—poor concordance; 0.5–0.75—medium concordance; 0.75–0.9—high concordance; 0.9–1—very high concordance. Concordance was also measured using Kappa weighted with linear and quadratic weights. Landis and Koch’s assessment criteria were adopted [25]: <0.20—poor concordance; 0.20–0.40—satisfactory concordance; 0.41–0.60—moderate concordance; 0.61–0.80—substantial concordance; >0.80 almost perfect concordance. For all statistical analyses, the significance threshold was set at p < 0.05. The analyses were conducted using IBM SPSS Statistics 29 (IBM Corp., Armonk, NY, USA, 2024) and JASP 0.18.3 (JASP Team, University of Amsterdam, Amsterdam, Netherlands, 2024). SPSS was used for descriptive statistics, reliability analyses, and correlations, while JASP was employed specifically for confirmatory factor analysis, as this functionality is not available in the standard version of SPSS.

3. Results

The descriptive statistics of the individual CLASS-S domains and dimensions are presented in

Table 2. Mean scores, standard deviation, skewness, kurtosis, and minimum and maximum values of individual CLASS-S domains and dimensions.

Item (Scale 1–7)	M	SD	Skew	Kurt	Min	Max
Emotional Support	3.75	0.79	−0.62	0.23	1.75	5.38
Positive Climate	4.50	1.10	−0.66	−0.04	1.63	6.38
Teacher sensitivity	4.83	1.17	−0.44	−0.37	2.13	6.75
Regard for adolescent perspective	1.92	0.56	2.44	10.37	1.13	4.63
Classroom Organisation	6.19	0.74	−1.36	2.34	3.50	7.00
Behavioural management	6.15	1.06	−1.84	3.66	2.13	7.00
Productivity	5.51	1.19	−0.81	0.21	2.13	7.00
Negative climate (reversed)	6.90	0.26	3.58	13.52	5.62	7.00
Instructional support	1.91	0.37	0.57	0.62	1.18	2.95
Instructional learning formats	3.46	0.98	0.18	−0.65	1.50	5.50
Content understanding	2.88	0.70	−0.02	−0.24	1.25	4.38
Analysis and inquiry	1.03	0.10	3.80	13.72	1.00	1.50
Quality of feedback	1.15	0.30	3.03	10.30	1.00	2.50
Instructional dialogue	1.01	0.06	5.53	33.00	1.00	1.38

. The highest score was given to the classroom organisation domain (M = 6.19), indicating very good classroom organisation. The “Negative Climate” dimension received the highest score, indicating a very low presence of negative teacher–student interactions. Both this indicator and ‘Behaviour Management’ have a strong negative skewness, suggesting a preponderance of high scores. The emotional support domain received a medium score (M = 3.75), which can be interpreted as a moderate level of emotional support. However, it is worth noting the very low score for the Regard for Adolescent Perspective (RAP) indicator, which simultaneously has a very high positive skewness and kurtosis, indicating a flattened and clearly asymmetrical distribution of scores. The instructional support domain achieved the lowest mean score (M = 1.91), which can be interpreted as a low level of interactional quality. Particularly low scores apply to the Instructional Dialogue and Analysis and Inquiry indicators, both of which have extremely high positive skewness and kurtosis, indicating that this type of support is extremely rarely observed in the recorded lessons.

Table 3. Correlations between domains and dimensions of CLASS-S.

CLASS-S Items	1	2	3	4	5	6	7	8	9	10	11	12	13	14
Domains
1. ES	---	0.51 **	0.65 **	0.93 **	0.93 **	0.47 **	0.50 **	0.37 **	0.59 **	0.68 **	0.59 **	0.32 **	0.27	0.09
2. CO		---	0.52 **	0.43 **	0.59 **	0.09	0.93 **	0.91 **	0.55 **	0.54 **	0.59 **	0.11	0.00	0.03
3. IS			---	0.59 **	0.59 **	0.37 **	0.45 **	0.48 **	0.37 **	0.95 **	0.92 **	0.22	0.68 **	0.51 **
Dimensions
4. PC				---	0.83 **	0.26	0.41 **	0.31 *	0.59 **	0.62 **	0.53 **	0.10	0.29 *	0.18
5. TS					---	0.24	0.61 **	0.42 **	0.58 **	0.63 **	0.58 **	0.28	0.20	0.09
6. RAP						---	0.05	0.09	0.14	0.34 *	0.29 *	0.57 **	0.32 *	0.02
7. BM							---	0.71 **	0.60 **	0.47 **	0.52 **	0.12	0.05	0.01
8. P								---	0.26	0.50 **	0.55 **	0.08	0.00	0.03
9. NC a									---	0.38 **	0.39 **	0.09	0.11	0.06
10. ILF										---	0.82 **	0.13	0.54 **	0.39 **
11. CU											---	0.19	0.52 **	0.35 **
12. AI												---	0.12	0.00
13. QF													---	0.80 **
14. ID														---

Note. CLASS-S—Classroom Assessment Scoring System–Secondary; ES—emotional support; CO—classroom organisation; IS—instructional support; PC—Positive Climate; TS—Teacher Sensitivity; RAP—Regard for Adolescent Perspectives; BM—Behaviour Management; P—Productivity; NC—Negative Climate; ILF—Instructional Learning Formats; CU—Content Understanding; AI—Analysis and Inquiry; QF—Quality of Feedback; ID—Instructional Dialogue; ^a—Reversed scale. * p < 0.05. ** p < 0.01. The correlation between the dimension and the respective domain is underlined.

shows the correlation coefficients between the three main domains of the CLASS-S tool: emotional support (ES), classroom organisation (CO), and instructional support (IS), as well as between the individual dimensions included in these domains. The results confirm the relevance of the structure of the CLASS-S tool—the dimensions follow the theoretical structure and are strongly related to their domains. Furthermore, correlations between the different domains are also noticeable, which may indicate that different aspects of the quality of teacher–student interaction are mutually reinforcing.

For the purposes of this study, CFAs were carried out for four different approaches. Presented in

Table 4. Factor loads and matching indicators of different models of CLASS-S.

	Three-Factor Model [12]			Single Factor Model	Two-Factor Model		Revised Three-Factor Model
	ES	CO	IS	ET	ES-IS	CO	ES	CO	IS
PC	0.85			0.74	0.70		0.84
TS	0.98			0.77	0.71		1.00
RAP	0.25			0.34	0.36
BM		0.98		0.62		0.98		0.93
P		0.71		0.57		0.72		0.76
NC a		0.61		0.54		0.60
ILS			0.94	0.89	0.92				0.92
CU			0.88	0.86	0.87				0.89
AI			0.18	0.21	0.20
QF			0.57	0.46	0.54
ID			0.43	0.33	0.39
χ2	154.252 (41)			216.895 (44)	191.881 (43)		14.040 (6)
χ2/df	3.76			4.93	4.46		1.158
SRMR	0.14			0.15	0.15		0.028
RMSEA	0.24			0.28	0.26		0.056
CFI	0.69			0.51	0.58		0.996
TLI	0.57			0.39	0.46		0.988

Note. CLASS-S—Classroom Assessment Scoring System–Secondary; ES—emotional support; CO = class organisation; IS = instructional support; ET = Effective Teaching; PC—Positive Climate; TS—Teacher Sensitivity; RAP—Regard for Adolescent Perspectives; BM—Behaviour Management; P—Productivity; NC—Negative Climate; ILF—Instructional Learning Formats; CU—Content Understanding; AI—Analysis and Inquiry; QF—Quality of Feedback; ID—Instructional Dialogue; ^a—Reversed scale.

, the CFA results show a poor fit of the data to the original three-factor structure of CLASS-S [12]. A similar poor fit is encountered with the one-factor structure and the alternative two-factor structure, representing the ideas of responsive teaching. In a modified model taking into account the specificity of PE classes, and after excluding several dimensions, we obtained a six-dimensional structure consisting of three domains, whose data fit indices are at a high level (χ²/df = 1.158; p < 0.05; SRMR = 0.028; CFI = 0.996; TLI = 0.998).

Figure 1 presents the modified CLASS-S model graphically. Reviewing the values of the standardised factor loadings, all dimensions show strong associations with the highlighted domain (>0.70) [26]. It was decided to add a residual covariance between the dimensions ‘teacher sensitivity’ and ‘behavioural management’. Referring to the CLASS-S theoretical construct, a teacher who is more sensitive to students’ needs can also better manage behaviour in the classroom, as he or she is able to respond to potential disruptions before they develop into larger problems. Therefore, the addition of a residual covariance between these indicators is justified.

In the second step of the CLASS-S validation, the reliability of the scales was examined using Cronbach’s alpha index, composite reliability (CR), and the average variance extracted (AVE).

Table 5. Reliability analysis results of CLASS-S.

Scale	CR	α	AVE	ES	CO	IS
Emotional Support	0.92	0.91	0.85	(0.92)
Class Organisation	0.84	0.82	0.70	0.49 ***	(0.84)
Instructional Support	0.9	0.87	0.83	0.65 ***	0.57 ***	(0.91)

Note. ES—emotional support; CO = class organisation; IS = instructional support CR = Composite reliability (desired value ≥ 0.75); α = alfa Cronbach (desired value ≥ 0.75); AVE = mean extracted variance (desired value ≥ 0.50). *** p < 0.001.

shows these coefficients together with Pearson’s r correlation coefficients between the scales and the square root of the diagonal AVE.

The results of the analysis indicate good reliability of the tool. The square root of the AVE was greater for each scale than the coefficients of its correlation with other scales. All scales were significantly and positively correlated with each other—the strength of the relationship was moderate for the correlation of emotional support with classroom organisation and strong for the correlation of emotional support with learning support and the correlation of learning support with classroom organisation.

The quality of teacher–student interaction according to the CLASS protocol was assessed by two independent observers. The concordance of their ratings is presented in

Table 6. Concordance of ratings by two independent observers (inter-rater reliability) on individual CLASS-S dimensions in each observation cycle (N = 200).

CLASS-S Dimensions	Cycles with Perfect Agreement (%)	Cycles with Agreement Within 1 Point (%)	ICC [95% CI]	Linearly Weighted Kappa	Quadratically Weighted Kappa
PC	60.5%	98.5%	0.930 [0.907–0.947]	0.718	0.868
TS	64.0%	94.0%	0.928 [0.905–0.946]	0.736	0.866
BM	73.5%	96.5%	0.938 [0.918–0.953]	0.748	0.883
P	64.5%	98.0%	0.946 [0.928–0.959]	0.760	0.896
ILF	52.0%	97.5%	0.892 [0.857–0.918]	0.617	0.804
CU	54.0%	98.5%	0.831 [0.776–0.872]	0.529	0.710

Note. PC = Positive Climate; TS = Teacher Sensitivity; RAP = Regard for Adolescent Perspectives; BM = Behaviour Management; P = Productivity; NC = Negative Climate; ILF = Instructional Learning Formats; CU = Content Understanding; ICC = intraclass correlation coefficient; 95% CI = 95% confidence interval.

. Observers were in almost 100% agreement on the Instructional Dialogue dimension. Above 60%, their ratings were in perfect agreement in four dimensions, while in two dimensions, full agreement was over 50%. In the CLASS assessment method, scores of ±1 point are considered acceptable and interpreted as concordance [12]. Hence, the table shows concordance taking into account the one-point margin. Consistency should be >80%. In this case, the concordance of the scores was between 94% and 100%. Thus, it can be concluded that the two independent observers were congruent in their ratings of teacher behaviour. This is also confirmed by the intraclass correlation coefficient (ICC). There was very high agreement in the dimensions “positive climate”, “teacher sensitivity”, “behaviour management”, and “productivity”. In the other dimensions, the ICCs indicated high concordance; in the case of the dimension “teaching dialogue”, concordance was medium. The concordance of the assessments measured by weighted Kappa indicated satisfactory to substantial concordance (linear weighting) and moderate to substantial concordance with quadratic weighting. Overall, the concordance scores of the two observers indicate adequate reliability of the modified CLASS-S instrument for Polish conditions and Physical Education settings.

4. Discussion

A review of the literature showed that CLASS-S had only been validated in non-US settings in the UK, Norway, Finland, and Mexico. In order to be able to apply this tool in Poland, there was a need for reliable validation. There are two practices in the literature for analysing the factor structure of this tool. The first, more commonly used, involves taking the averaged scores of the individual dimensions of a single teacher assessed by multiple observers for analysis, which are then aggregated to obtain the scores of the three domains. In accordance with such a procedure, validations have been performed by Allen et al. [19], Pianta et al. [12], Virtanen et al. [14], Westergård et al. [16], Hafen et al. [20], and the authors of the present study. However, there are claims [13] that a better method for validating CLASS-S is hierarchical analysis, which allows the factor structure to be modelled at different levels and thereby reveal more complex relationships. On the other hand, analysis hierarchies are more useful for discovering new structures than for confirming established theoretical models.

Results from 24 Polish schools suggest that the CLASS-S with some modifications has satisfactory psychometric properties. The hypothesised three-factor structure proved to be the best fit to the collected research data compared to the one- and two-factor structures. However, the adopted model required significant modifications. A good fit required the removal of five dimensions, as only then were acceptable indicators obtained. However, it was established that the CLASS-S tool, although it assesses teacher–student interactions, does not match the nature of all types of lessons. To date, no one has used CLASS-S to assess the quality of interactions in PE lessons, and so the attempt made here was a world first. It turned out that under Polish [education] conditions, the original CLASS-S structural model did not fit the collected data. At the same time, this could have been due to either the specifics of PE classes or the specifics of the Polish education system. This cannot be resolved at this point, as there were no observations of teacher–student interactions during other school subjects in secondary school in Poland. We do have a reference to the use of the Classroom Assessment Scoring System—Upper Elementary [17] in the SUEK study [27] in Poland, but the team of researchers did not use the CLASS protocol directly, rather they built a computer application, ARGOS, based on it, supporting the process of direct observation. At that time, the recommended assessment from video footage was not opted for, and only some results were published, but without validation of the tool. The observers in the SUEK study were also not trained at Teachstone, but in Poland for their own use. It is, therefore, difficult on the basis of the SUEK study to assess the usefulness of CLASS-UE in a Polish school. Assuming that the need to modify CLASS-S was related to the specificity of the Polish education system, the removal of four important dimensions from the model, “taking into account the student’s perspective”, “analysing and solving problems”, “quality of feedback”, and “teaching dialogue”, was the result of basing teaching on the traditional Herbartian approach to education [28]. Despite the fact that in the declarations of educational authorities and public school teachers themselves, the Polish school [system] is moving away from this approach, based only on this one study, it can be assumed that it is still oriented towards one-way transmission, rigid adherence to the timetable, and a lack of flexibility, discussion, exploration, and problem-solving. The idea of responsive teaching, i.e., one in which the teacher follows the student, tuning in to the student’s current needs, interests, ideas and expressed ideas, has not yet penetrated the Polish school system [29,30]. As a result, the present study has shown that during PE in Polish schools, the student is “silent”. Teachers conduct one-sided communication; they order exercises and expect implementation. If they already ask something, they do not expect an answer. They do this because they believe that it is the task of ‘mental’ subjects to impart knowledge, while that of PE is to develop and shape the body. For this purpose, dialogue with the student seems less useful. Teachers have always been taught about ways to intensify activities and less about the quality of verbal communication.

The confirmatory factor analysis conducted showed only the best-fit model for the data. This should not be interpreted as meaning that teacher–pupil interactions in these dimensions should not characterise interactions during PE lessons/sports activities. The conclusion is the opposite. If teachers and the Polish school system want to achieve better educational outcomes, a space for possible improvement has been identified. A student in a Polish school should not be ‘dumbed down’. This is indicated by research linking high-quality teacher–student interaction to educational achievement [12].

The ‘instructional dialogue’ dimension removed in the Polish adaptation, even in observations of ‘theoretical’ lessons in other countries, was rated low in all studies where it was considered [14,15,16]. In the observations of Polish PE teachers, it occurred incidentally, as did prompting students to analyse and solve problems. Only in the UK [19] were teachers rated at an average level in this aspect. Polish teachers were also rated low in the “Quality of Feedback” dimension, indicating that the use of feedback in their interactions with students was limited. By far, teachers in other countries used this technique more frequently (medium ratings, except for a low rating in the Mexican study) [15].

Interestingly, PE lessons in Poland were rated highest in the negative climate dimension. The score of 6.90, close to the maximum, should be interpreted as the absence of negative emotions during lessons, and rare episodes of anger, upset, or physical aggression of the teacher or students. No disrespectful, ironic, or demeaning messages were observed. This high rating can be explained in two ways; the first is very optimistic and is indeed related to the absence of a negative climate in the lesson. The objectives of PE prescribe making physical activity a source of positive experiences. On the other hand, we may be dealing with the ‘observer effect’; the teachers knew that the lesson was being recorded and could modify and correct their own behaviour. Perhaps the lessons did not reflect the real and ‘normal’ atmosphere during the lesson. This is a known and accepted flaw in the observation method. Whatever the determinants of this assessment, it did not differentiate between teachers and was, therefore, not included in the Polish modification of the CLASS-S tool. The negative climate was also removed in the Mexican validation. Razo and Peralta [15] explained this by cultural differences that may affect the factor structure of diagnostic tools. They considered that, for example, discipline and order may be culturally conditioned. In Mexico, episodes of negative student attitudes towards teachers are rare. This is due to enforced subordination and top-down discipline. In a Mexican school, there is no asking of the student, but a demand for commitment, obedience, and the achievement of expected results. Such an interpretation is incompatible with Polish conditions. It is more likely to be society’s current expectations of teacher behaviour. It is simply that certain behaviours are not accepted in a Polish school.

The dimension ‘Regard for Adolescent Perspective’ was also not included in the final adopted CLASS-S model. The reason for the exclusion in this case was not related to the lack of observation of teacher–student interaction in this dimension, but to the lack of correlation of this variable with the other items, the emotional support domain, and the construct as a whole. The mean score of teachers on this dimension was 1.92 points (on a scale of 1–7). It is worth discussing this score, as it informs us that during PE lessons, teachers were not flexible, and did not give space for students’ ideas or their own initiative. Moreover, according to the description of this dimension in CLASS-S, a low score also means that the content of the lessons was not related and linked to the student’s everyday life and so, the usefulness of this content in out-of-school settings was not communicated to the students. Finally, students were not allowed to influence the course of the lesson, or make choices, were not given the chance to take responsibility for the course of their education, and were not given the opportunity to take leadership roles. The average grade was closer to 2 points rather than 1 point, probably because the specifics of PE include performing tasks in groups (primarily in team games). As the CLASS-S has not been used before to assess interaction in physical activity settings, our own results can only be compared to those obtained from observations of lessons taking place in a ‘traditional’ classroom. Not all studies included this dimension (RAP). Virtanen et al. [14] also excluded ‘taking into account the student’s perspective’ from their model. They explained the low discriminatory ability of this dimension by the high capacity of the term. They argued that it encompasses not only emotional support interactions but also teacher practices related to classroom organisation (e.g., allowing students to choose tasks, giving them responsibility and leadership opportunities) and learning support (e.g., encouraging students to think, share ideas, putting them in problem situations). Consequently, it is difficult to attribute this dimension solely to the domain of emotional support. Rather, it concentrates elements of the three domains, i.e., teacher flexibility, consideration of student autonomy and freedom of action, and teacher sensitivity to their individual needs. In other studies, this dimension had statistical and practical utility. In three American studies, students were assessed at a mean level ranging from 2.86 in the UTQ-A study to 3.49 in the Secondary MTP study, which is interpreted as the borderline of low to medium interaction quality. Teachers were rated similarly in studies outside the USA.

The validated CLASS-S tool in a modified version adapted to the specifics of Polish PE lessons is a promising tool for supporting the achievement of Sustainable Development Goals, especially SDG 3 and SDG 4. As the results indicate, high scores on the emotional support domain (e.g., positive climate, teacher sensitivity) correspond with the SDG 3 assumptions promoting mental well-being and stress reduction in students, which is particularly important in the context of the deteriorating health status of Polish youth noted in the reports of the Ministry of Development and Technology [5,6]. At the same time, the low scores on the dimension ‘taking into account the student’s perspective’ suggest the need for greater flexibility in teacher–student dialogue, key to achieving SDG 4, which envisages inclusive and co-determined education. These results are consistent with the findings of Baena-Morales et al. [6], who highlighted that the quality of teacher–student interaction directly translates into the inclusiveness of the educational process (SDG 4) and the formation of health-promoting attitudes (SDG 3). However, as in the study by Royet et al. [31], the observed deficits in instructional support (e.g., low levels of teaching dialogue) point to the need for further modifications of the tool to better support critical thinking and student engagement—key competencies for sustainability. A practical conclusion is to recommend the inclusion of CLASS-S in PE teacher training programmes, which may contribute to better integration of the SDGs in PE, especially in the context of a holistic approach to health (SDG 3) and addressing exclusion (SDG 4).

This study has several limitations that need to be taken into account when interpreting the results. Firstly, the sample was limited only to PE lessons in Polish secondary schools, which may limit the possibility of generalising the results to other school subjects or other educational stages. Secondly, the observations relied on video recordings, which may have influenced teachers’ behaviour (the so-called camera effect). Thirdly, the applied modification of the structure of the CLASS-S tool, although adapted to Polish [education] conditions, makes direct comparisons with the results of studies conducted in other countries difficult. Finally, the study was based solely on the analysis of interactions by external observers, without including the perspective of the students and/or teachers themselves, which could enrich the interpretation of the data.

5. Conclusions

The data collected for this research indicate that the modified CLASS-S is a valid and reliable tool for assessing the quality of teacher–student interaction during PE lessons. The model demonstrated excellent fit (χ²/df = 1.158; CFI = 0.996; TLI = 0.998), with inter-rater reliability ranging from 94% to 100% (ICC = 0.928–0.946). This can be considered an important contribution to Physical Education methodology as it opens up a discussion on how to improve the effectiveness of PE. In addition to statistical analyses confirming one or other internal structures of CLASS-S, teacher–student interaction can be approached from the practical side. Just pointing out the determinants of student achievement, mediated by a high-quality relationship with the teacher, from a practical point of view is already cognitively interesting. Showing PE teachers a space in which they can improve themselves can be a stimulus for their decision to develop further.

These outcomes offer practical implications for the education of teachers of Physical Education in Poland. The training programmes should put more focus on enhancing teachers’ social skills—specifically in fostering instructional dialogue, encouraging student independence, and encouraging problem-solving conversations. This would enable future PE teachers to shift from one-way teaching and develop more inclusive, active learning environments. Also, countries having education systems as equally traditional and teacher-oriented as in Poland can be informed by the present findings. The case of Poland demonstrates how it is possible for cultural and system-level factors to influence teacher–student interactions, which can be observed and impact the applicability of internationally established evaluation instruments. This research can thereby be of use in an international debate about the cultural adaptation of classroom observation frameworks as well as in favour of evidence-based reforms of teacher education in various national contexts.