Teachers implement different types of instruction, and these differences have significant consequences for the quality of instruction and how students learn science (Aditomo & Klieme, 2020; Mikeska et al., 2017). Nevertheless, studies examining the different patterns of teachers’ instructional practices in science are scarce. Little is known about distinct patterns of science practices and how they may relate to student cognitive and affective outcomes. Even fewer studies have compared the nature of these instructional practices across educational stages. Knowledge about teachers’ patterns of instructional practices and their relationships to student outcomes is beneficial for policy and practice to support teachers in implementing science instruction that address the needs of diverse learners.

International large-scale assessments (ILSAs) like the Trends in International Mathematics and Science Study (TIMSS) offer an opportunity to study the profiles of instructional practices in science classrooms. In fact, TIMSS is the only ILSA study that assesses not only students’ cognitive but also affective outcomes in primary and secondary education. TIMSS also assesses various science teaching and learning aspects, including students’ perceptions of instructional quality, student characteristics, and teacher competence (Martin & Mullis, 2016). Since TIMSS includes a nationally representative sample of students, findings about the profiles of instructional quality derived from TIMSS data may provide the potential for generalizability to inform policy and practice at the national level. This study may be also beneficial for researchers who are unfamiliar with TIMSS, as it offers insights into how TIMSS data could be utilized to advance research on instructional quality.

Consequently, the present study utilizes TIMSS 2015 data to investigate the different patterns of instructional quality and the extent to which these patterns may differentiate student outcomes. Although a substantial body of research on the relationships between teacher instruction and student outcomes exists (Aditomo & Klieme, 2020; Mikeska et al., 2017; Nilsen, Scherer, & Blömeke, 2018), current studies have typically applied a variable-centred approach by examining an entire sample or population without considering the possibility that these relationships may differ across subgroups of teachers (e.g., Areepattamannil, Cairns, & Dickson, 2020; Teig, Scherer, & Nilsen, 2018). Teachers may implement different practices in the classrooms based on their teaching beliefs and philosophy along with their previous education and experience (e.g., Bonneville-Roussy, Bouffard, Palikara, & Vezeau, 2019; Teig, Scherer, & Nilsen, 2019). Using a person-centred approach, the current study instead identifies subgroups of teachers who demonstrate unique profiles of instructional quality. This approach allows researchers to investigate the antecedents of these profiles on the one hand and the profile-specific effects on student outcomes on the other hand.

Since teaching and learning activities are usually bound to a multilevel context in which students are nested within classrooms (Flunger et al., 2019), it is possible that the qualitative differences in the profiles of instructional quality may largely depend on the characteristics in the classrooms (e.g., teacher background and beliefs). In the same classroom, students’ perceptions of teacher practices may not independent of one another, and this dependency should be modelled when applying person-centred methods (Flunger et al., 2019; Henry & Muthén, 2010). Thus, this study applies multilevel latent profile analysis (MLPA) that takes into account the variability of perceived instructional quality at the student- and classroom-levels.

Specifically, the present study examines teachers’ profiles of instructional quality in Grades 5 and 9 to represent primary and secondary education, respectively, with the following research questions (RQs):

1. What kinds of patterns can be identified from the profiles of instructional quality at the student- and classroom-levels?
2. To what extent do science cognitive and affective outcomes differ across the profiles of instructional quality at the student- and classroom-levels?

We utilize TIMSS 2015 data from representative samples of primary and secondary science classrooms (i.e., Grades 5 and 9) in Norway. Students were asked about practices related to clarity of instruction (e.g., my teacher is good at explaining science), engaging teaching (e.g., my teacher gives me interesting things to do), social and emotional support (e.g., my teacher listens to what I have to say), and subject domain support (e.g., my teacher tells me how to do better when I make a mistake). Students’ perceptions of teacher practices were used to derive the profiles of instructional quality. Student cognitive outcome was estimated via a measurement model that produced a set of five plausible values to represent the range of student achievement. To measure student affective outcome, we used one item on science motivation that asked students to rate their agreement with the following statement, “I enjoy learning science”, ranging from 0 = disagree a lot to 3 = agree a lot. We employed MLPA (Henry & Muthén, 2010; Vermunt, 2008) to identify profiles of instructional quality at the student- and classroom-level in Grades 5 and 9 (RQ1). Since MLPA is explanatory in nature, we had no priori about the number of latent profiles in both levels. Hence, we performed a series of analyses with increasing numbers of latent profiles and compared the resultant models in order to derive well-defined profiles with substantive interpretation. Next, we examined whether the mean of science achievement and motivation differed across the profiles at the student- and classroom-levels (RQ2). For the models that predicted science achievement as a distal outcome, the analyses were conducted five times (i.e., once for each of the plausible-values datasets), and the resultant model parameters were pooled following Rubin’s combination rules (Rubin, 1987) using the syntax TYPE=IMPUTATION in Mplus.

The present study adds to research on instructional quality by investigating whether subgroups of teachers demonstrate unique teaching profiles that emphasize clarity of instruction, engaging teaching, social and emotional support, and subject domain support. In Grade 5, we identified five profiles with distinct patterns: the lowest quality, the lower quality with teachers’ good explanation of science, the medium quality with teachers’ good explanation of science but lower opportunity for students to show their learning, the medium quality with poor teachers’ explanation of science, and the highest quality. In Grade 9, we identified four profiles: the highest, moderately high, moderately low, and lowest quality. The findings also indicate that science motivation and, to some extent, achievement are significantly different across the profiles of instructional quality at the student- and classroom-levels in both grades. This study is among the first to investigate unique profiles of instructional quality in science, both in primary and secondary education. Given that current research investigating instructional quality has mostly focused on a specific grade and level of analysis, this study contributes to the existing literature by comparing how the patterns and their relations to student outcomes vary, specifically in science teaching. Using a person-centred approach, this study offers an alternative perspective to model instructional quality by analysing how the four aspects of instruction are organized within teachers in primary and secondary science classrooms. Educational policies that aim to improve instructional quality often take a one-size fits all approach, such as through interventions in teacher education and professional development. In contrast, this study shows that science teachers cluster into distinct teaching profiles that also vary between the educational stages. By focusing on individual teachers, we can design specific interventions to target different aspects of instructional quality that these teachers urgently need to improve.

Aditomo, A., & Klieme, E. (2020). Forms of inquiry-based science instruction and their relations with learning outcomes: Evidence from high and low-performing education systems. International Journal of Science Education, 1-22. Areepattamannil, S., Cairns, D., & Dickson, M. (2020). Teacher-Directed Versus Inquiry-Based Science Instruction: Investigating Links to Adolescent Students’ Science Dispositions Across 66 Countries. Journal of Research in Science Teacher. Bonneville-Roussy, A., Bouffard, T., Palikara, O., & Vezeau, C. (2019). The role of cultural values in teacher and student self-efficacy: Evidence from 16 nations. Contemporary Educational Psychology, 101798. Flunger, B., Trautwein, U., Nagengast, B., Lüdtke, O., Niggli, A., & Schnyder, I. (2019). Using Multilevel Mixture Models in Educational Research: An Illustration With Homework Research. The Journal of Experimental Education, 1-28. Henry, K. L., & Muthén, B. (2010). Multilevel latent class analysis: An application of adolescent smoking typologies with individual and contextual predictors. Structural Equation Modeling, 17(2), 193-215. Martin, M. O., & Mullis, I. V. (2016). TIMSS 2015 assessment frameworks. Chestnut Hill, MA: TIMSS & PIRLS International Study Center, Boston College and International Association for the Evaluation of Educational Achievement (IEA). Mikeska, J. N., Shattuck, T., Holtzman, S., McCaffrey, D. F., Duchesneau, N., Qi, Y., & Stickler, L. (2017). Understanding science teaching effectiveness: Examining how science-specific and generic instructional practices relate to student achievement in secondary science classrooms. International Journal of Science Education, 39(18), 2594-2623. doi:10.1080/09500693.2017.1390796 Nilsen, T., Scherer, R., & Blömeke, S. (2018). The relation of science teachers’ quality and instruction to student motivation and achievement in the 4th and 8th grade: A nordic perspective. In Nordic Evaluation Network (Ed.), Northern Lights on TIMSS and PISA 2018 (pp. 61-94). Denmark: NCM. Rubin, D. B. (1987). Multiple imputation for nonresponse in surveys. New York: John Wiley & Sons. Teig, N., Scherer, R., & Nilsen, T. (2018). More isn't always better: The curvilinear relationship between inquiry-based teaching and student achievement in science. Learning and Instruction, 56, 20-29. Teig, N., Scherer, R., & Nilsen, T. (2019). I know I can, but do I have the time? The role of teachers’ self-efficacy and perceived time constraints in implementing cognitive-activation strategies in science. Frontiers in psychology, 10(1697). Vermunt, J. K. (2008). Latent class and finite mixture models for multilevel data sets. Statistical Methods in Medical Research, 17(1), 33-51.