Dialogical modelling processes: conversations for the social construction of scientific models in the science classroom

Abstract

Classroom dialogue is fundamental in a classroom guided by socio-constructivist views of learning due to the learning experiences teachers can provide their students through them. This descriptive and exploratory case study aims to analyse the discourse that enables teachers to facilitate students' expression of models through the study of a phenomenon in a science class. The case study presents the experiences of four secondary school chemistry teachers who implement the same teaching and learning sequence about chemical combinations. The classes were recorded, transcribed, and analysed using a qualitative research framework. The results show that teachers facilitate learning opportunities through intermittent spaces of productive classroom conversation. The analysis of these dialogical spaces lets us identify a highly distinctive dialogic process that permeates instructional modelling opportunities, which focus on four dimensions of the management of the modelling conversation: teachers' conversational directionality, the presence of non-productive classroom conversation, moments that trigger the teacher's discursive decisions and the construction of the model through instructional modelling cycles.

---

Introduction

From the perspective of model-based science education, learning science at school involves the construction of increasingly sophisticated mental representations to explain phenomena relevant for the construction of a citizenship able to make socially and environmentally responsible decisions (Marzábal et al., 2024). The evolution of students' models, produced through constructing and reconstructing their mental representations, is known as the modelling process (Gouvea and Passmore, 2017; Oliva, 2019).

In order to promote this modelling performance in the classroom, a set of phases has been proposed as useful to guide science teaching and learning sequences. These comprise the modelling cycle, which includes the phases of familiarisation, discussion and consensus (Garrido and Couso, 2024). In the familiarisation phase, the approach of a contextualised problem situation is expected to facilitate students' recognition of the necessity for a model, its elicitation and its expression. The latter makes the presence of different versions of the model in the classroom visible. The discussion phase is associated with asking questions for students to express and use their ideas in an iterative process that encourages them to test and contrast their ideas to evaluate and revise their initial model, leading to its refinement. Finally, in the consensus phase, a final consensual model emerges as is agreed upon through negotiation and structuring processes that allow for the progressive convergence of the students' ideas towards a collective understanding.

The modelling cycle is based on the socio-constructivist perspective of learning, which acknowledges the social character of the construction, evaluation and adjustment of models and the essential role of teachers in fostering teaching environments that facilitate these model-construction processes (Gobert and Buckley, 2000). In each phase of the modelling cycle, teachers engage in actions designed to promote student involvement and participation in the joint construction of models through social interaction (Gilbert and Justi, 2016).

Given the social nature of the modelling process, classroom interaction, particularly conversation, plays a central role (Windschitl et al., 2012; 2018; Gray and Rogan-Klyve, 2018). Classroom conversation is recognised as the verbal interactions between the teacher and students, managed by the teacher during the lesson (Soysal, 2019). These interactions are discursive in nature but also instructional. To achieve modelling purposes, it is not enough to provide interactive and dialogic conversational spaces (Mortimer and Scott, 2003). Rather, teacher interventions must also have a pedagogical intentionality that generates opportunities for students to test their ideas to transform them adequately through interaction with others (Chin, 2007; Téllez-Acosta et al., 2023).

The existing literature identifies such interactions as productive classroom conversations (Michaels et al., 2009; Michaels and O’Connor, 2015). Through these interactions, the teacher is expected to encourage students to test their ideas and experiences and construct progressively more sophisticated ones through teacher-specific talk moves (Chin, 2007; Michaels and O’Connor, 2015; Park et al., 2017). Soysal (2019) has proposed a set of indicators to recognise the dialogues that effectively mediate this collective model's construction. These indicators refer to (a) elaborated intellectual exchanges between peer communities, (b) rigorous negotiation of different points of view, (c) the incorporation of epistemic responsibility in the conversation, (d) synchronisation of interactions during the conversation for meaningful learning, and (e) the discursive role of the teacher. In the view of the author, the presence of these indicators enables the identification of productive classroom conversations, which are dialogues that seek to facilitate learning through a collective process in which ideas are shared and discussed freely to build a cumulative understanding from the milieu of ideas and rationalities present in the classroom (Alexander, 2018).

From our point of view, a productive conversation would allow the mediation of the modelling process. From a socio-constructivist perspective, mediation constitutes a central pedagogical responsibility, as teachers strategically orchestrate discourse to promote idea development and assist students in articulating their thinking (Palincsar, 1998). Accordingly, teachers should use students' contributions to establish the foundation for forthcoming discussions. This can be achieved by opening or closing the dialogic spaces in the learning processes, considering the ideas presented by students while deciding how to include them in the conversation, thereby facilitating the development of scientific ideas (Lehesvuori et al., 2013). Modelling processes thus necessitate a specific discourse that encourages the expression, evaluation and refinement of students' initial models in a process to collectively enrich them (Merino and Izquierdo, 2011; Guy-Gaytán et al., 2019). Given the former, how teachers facilitate classroom discourse influences the modelling process, as it determines the direction and continuity of the processes of co-construction of meaning in each phase (Hennessy et al., 2016; Hamnell-Pamment, 2023).

The study of modelling dialogue has been approached from several different perspectives. These include the examination of metamodelling conversations (Gray and Rogan-Klyve, 2018), the implications of discourse in a modelling classroom (Guy-Gaytán et al., 2019), the study of discursive strategies to construct models (Vergara et al., 2025), and scaffolding strategies in enquiry (Kim, 2020). Although theoretical correspondences between the phases of the modelling cycle and the teaching practices associated with conducting the modelling classroom conversation can be anticipated, this articulation has yet to be fully explored. This justifies the need for further investigation into the communicative opportunities provided by teaching practices for student learning and how these articulate modelling-oriented science teaching in the school context (Shi et al., 2021).

In this study, we aim to contribute to this line of research by studying dialogical interactions in the science modelling classroom. In order to achieve this, we will address the following research question:

How does teacher management of productive conversation in the science classroom shape the instructional modelling processes?

With the aim of answering this question, an exploratory case study was conducted, focusing on the dialogic classroom conversation of four chemistry teachers who implement the same modelling-based teaching and learning sequence (TLS) co-designed by them and science education researchers. This has enabled us to characterise how their management of the classroom conversation shapes instructional modelling processes in different ways, which are responsive to the same design.

Methodology

Method

This study conducts a classroom discourse analysis (Johnstone, 2000) through a descriptive and exploratory case study approach (Yin, 2003), characterising dialogical interactions in the modelling-oriented science classroom. We analyse teachers' discourse and explore how they facilitate students' modelling processes through turns of speech at each phase of the modelling cycle.

Participant

This research analyses the conversation during chemistry lessons taught by four secondary school teachers in the Chilean education system (Table 1). The teachers in question participated in a community of practice focused on modelling as a teaching approach. In the context of the community, the teachers, with the mediation of one of the authors of this article, co-designed a modelling teaching and learning sequence focused on the model of chemical combinations. The objective of the TLS was to identify the chemical substances present before and after the reaction between different amounts of copper sulphate and iron, employing the ponderal laws. The deliberate sampling strategy (Woodsong et al., 2005) permitted forming a specific and information-rich group (Patton, 2015).

Table 1 Teachers’ characteristics

Teachers	Type of school	Years of experience	Students	Amount of students
Joseph	Public	7 years	First year of secondary school	41
Chris	Charter	11 years	First year of secondary school	42
Maria	Public	10 years	First year of secondary school	21
Caroline	Public	12 years	Third year of secondary school	19

---

The analysed classes correspond to the implementation of the jointly planned lesson, which they carried out in their respective educational settings. All teachers and students were informed about the purpose and scope of this study and voluntarily provided their consent to participate.

Context

The TLS was designed for students to identify the chemical species present before and after the reaction between different amounts of copper sulphate and iron, applying ponderal laws. The design takes into account the presence of four key ideas of the chemical change model according to Merino and Izquierdo (2011): (a) some substances disappear, and others appear; (b) mass and elements are conserved; (c) substances react in fixed proportions and (d) particles, atoms and bonds can represent chemical change.

The design of the TLS proposed the implementation of three experiments, each comprising an activity with associated tasks (Table 2). Teachers were afforded the autonomy to determine the implementation and segmentation of the TLS and make adaptations per their contextual requirements.

Table 2 Tasks with modelling potential that are implemented by teachers from the co-designed teaching and learning sequence

Design					Implementation
TLS	Activity	Tasks	TLS	Expected answers	Students work
Exp. 1	Initial observation of the reaction between copper sulphate and iron	T1: Represent and describe the initial state (reactant) and final state (product)
Exp. 2	Comparison of the reaction of different amounts of iron with a fixed amount of copper sulphate solution	T2: Represent and describe the final state after doing the reactions (macro)
		T3: What will be the changes in the reacting substances?
		T7: What do you think the final state of the reaction will be like?
		T8: Predict what is going to happen microscopically. What should you see?
Exp. 3	Comparison of the reaction of different amounts of copper sulphate solution with a fixed amount of iron	T4: Draw and explain what you think the final state of the reaction will be like.
		T5: Represent and/or describe the end state of the reaction
		T6: Represent the final state after doing the experiment (micro).

---

The criteria for designing tasks with modelling potential proposed by Cortés-Morales and Marzábal (2025) were employed to guarantee that the design was oriented towards modelling. According to the authors’ systematic review the modelling potential of a task is related to the fact that the task: (a) has an open question that enables a broad repertoire of acceptable responses or resolution strategies, (b) is situated in a phenomenological context, and (c) demands for the students engagement and enactment of a science practice to formulate a response to the question posed in the task. Any task that has these three characteristics can be considered modelling-oriented. An example can be found in Table 3.

Table 3 Characterisation of the modelling potential of an implemented task (T7) in this study

Task	Degree of openness	Context	Science practice	Modelling potential
In the reaction of 80 ml of copper sulphate solution (0.1 M) with 2 g of iron. What do you think the final state of the reaction will be like?	Open	Scientific	Prediction	Yes

---

After this process, it was established that these criteria were met in all the tasks implemented by teachers, as the tasks were open-ended, situated within a specific empirical context, and their demands required the deployment of a science practice that can trigger modelling processes (Cortés-Morales and Marzábal, 2025).

Strategies for data collection

The data for this qualitative study were obtained from video recordings of implementing a teaching and learning sequence. The video recordings were made following the analytical strategy proposed by Lehesvuori et al. (2018) to analyse teaching processes. This entailed the utilisation of a camera positioned at the back of the classroom to record the teacher's interactions with the whole group, as well as a record of the teacher's interactions with the smaller groups. The teacher is followed by the camera and microphone for their conversations. This recording allows for the acquisition of audio and video of conversations with one small group at a time and the whole group.

The video recordings of the teachers' lessons for Joseph, Chris and Maria are 90 minutes long, while the lesson for Caroline lasts 60 minutes. The data set comprises productive and non-productive classroom discourse on tasks associated with different activities. As the analysis was conducted on modelling-oriented tasks, the data were reduced by selecting the instances corresponding to classroom conversations between teachers and their students in tasks associated with constructing a school science model.

Subsequently, a second reduction of data was carried out in the tasks with modelling potential through the exclusion of classroom conversation, the purpose of which is external to the carrying out of the tasks. This entailed the exclusion of instances of conversation related to those that occur in moments of transition from one classroom to another, the distribution of materials, the management of the classroom climate, and other such occurrences that fall outside the scope of this study.

Due to those data reductions, the findings comprise the conversations while implementing the tasks with modelling potential, making the blocks of time less than the duration of the complete lessons. Those conversations are continuous in time, and the tasks are chronologically occurring in the implementation. In this manner, the data allude to visual registers that consider conversation a means for the social construction of a school scientific model through dialogue in tasks with modelling potential.

Strategies for data analysis

This study addresses the research question through a case study approach, employing a qualitative classroom discourse analysis. This analysis encompasses several instances of conversation occurring within the context of different tasks within a teaching and learning sequence. The data were collected from a single lesson, which four different teachers taught. The subsequent qualitative analysis attempts to integrate two frameworks, modelling and productive conversation, to analyse classroom discourse.

The dialogue between teachers and students, or between students engaged in a task with modelling potential, was transcribed using the ‘play script’ format (Johnstone, 2000) in order to highlight each communicative act of teachers and students as turns of speech. The previously reduced fragments were coded considering key aspects of the discourse, using already existing analytical tools:

(a) Tasks with modelling potential: the identification of the point at which each task with modelling potential is initiated by teachers (Cortés-Morales and Marzábal, 2025).

(b) Typology: the distinction between classroom (non-productive) and productive conversations (Soysal, 2019).

(c) Organisation: Teachers interact with the entire class or smaller groups (Alexander, 2018).

(d) Identification of the modelling cycle phase promoted by the teacher discourse: familiarisation, discussion and consensus (Garrido and Couso, 2024).

This analysis identifies the percentage of productive teacher discourse in each modelling phase, thus enabling comparison between the different phases and the characterisation of the idiosyncrasies in the use of discourse for modelling by the participating teachers.

In the analysis of the data, it was found that teachers’ productive conversation was not continuous, as it alternates interactive/dialogic dialogues with conversations that correspond to different communicative approaches, such as non-interactive/dialogic, interactive/authoritative or non-interactive/authoritative (Mortimer and Scott, 2003). Although research in productive classroom conversations is an area with extensive research, it lacks a term for moments when the productive conversation ends and another type of communicative approach starts. For the purposes of this study, every turn of speech that cannot be characterised as productive classroom conversation will be classified as “non-productive”. This typology seeks to refer to turns of speech that do not comply with the canonical definition of productive conversation, but which do not necessarily have to be unproductive in terms of educational or learning gains.

The identification of teachers who facilitate students' familiarisation, discussion and consensus of ideas to a greater extent, as well as those who present difficulties in this area, enables the construction of cases that describe and interpret the diverse ways in which productive conversation is used to promote student modelling. The discourse analysis carried out for each case is visualised using chronographs. A chronograph is a temporal representation of discursive processes, which enables the visualisation of the duration and progress of classroom conversation during a lesson on a chronological timeline.

In the chronography, various symbolic codes were employed to represent the key aspects of the discourse analysed (categories a, b, c and d), detailed in Table 4. An example of the coding process can be seen in Appendix 1.

Table 4 Codes and Keys used to analyse the discursive process and construct the chronography

Category	Subcategory	Descriptor	Representation
Time		Representation of the chronological time of the class	0	1	2	3	min
Task (Cat.a) (Cortés-Morales and Marzábal, 2025)		Beginning of each new task with modelling potential, i.e., the task has an open question that is situated in a context where a science practice must be undertaken.	*
Typology (Cat.b) (Soysal, 2019)	Non-productive classroom conversation	None of the indicators of productive conversation is identified in the conversation.
	Productive conversation	One or more indicators of productive conversation are present, i.e. (a) elaborated intellectual exchanges between peer communities, (b) rigorous negotiation of different points of view, (c) the incorporation of epistemic responsibility in the conversation, (d) synchronisation of interactions during the conversation for meaningful learning, and (e) the discursive role of the teacher.
Organisation (Cat.c) (Alexander, 2018)	Whole group	Discursive interactions between the teacher and the whole class	W.G.
	Small group	Discursive interactions between the teacher and a group of students	G.n*
			*Number of small group
Modelling Cycle (Cat.d) (Garrido and Couso, 2024)	Familiarisation	Familiarising with the phenomena, recognising the need for a model and using/expressing initial ideas	Axis of coordinates
	Discussion	Testing and contrasting students' ideas to check their adequacy, consistency and robustness based on observation/experimentation data and arguments from others
	Consensus	Negotiation, structuration, reorganisation and synthesis of the ideas discussed to express a consensus model that is progressively more aligned with the school science model.
Example
		Representation of the dialogue in the flow of a task

---

To illustrate these cases qualitatively and to facilitate interpretation, instances of authentic classroom discourse representative of the type of discourse most present in each case are selected and reproduced. The examples are extracts from conversations in which the teacher's and students' (S. no.) turns are differentiated in the context of the implementation of a particular task, which is associated with one of the modelling phases. However, it should be noted that some of the intentionality may be lost in the translation process.

The results section presents the quantitative analysis conducted, along with a description and interpretation of the four cases identified.

Methodological validity and reliability

Multiple strategies were employed to ensure a rigorous analysis and enhance the methodological validity of this study. First, data collection was conducted meticulously, incorporating high-quality video and audio recordings. Second, explicit criteria grounded in established theoretical frameworks guided analytical decisions, ensuring consistency and justification. Lastly, two researchers performed data reduction and analysis independently, resolving discrepancies through expert judgment and involving a third reviewer when necessary. These measures reinforced the study's transparency, reproducibility, and methodological rigour.

Additionally, the necessary ethical protocols were followed, ensuring participants were informed about the study and voluntarily agreed to participate by signing a consent form. Due to ethical confidentiality, the names of the teachers have been changed. For this, Human Subjects Research (HSR) protocols were followed accordingly to Social-Behavioural-Educational (SBE) norms, which are required by the ethics committee in our university, which is the institution that approved the study.

Results

The analysis made it possible to characterise the management of classroom conversation with different focal points, providing a more comprehensive and enriched view of discourse during a modelling-oriented lesson. The use of tasks with modelling potential gives rise to modelling processes in which it is possible to identify the modelling phases, how teachers organise and manage students' work, and instances of productive discussion and the speech turns involved. The study is conducted in two stages to achieve this. First, the quantitative aspects were examined to identify patterns in the teaching orchestration of productive conversation during modelling. This includes analysing the number of speech turns, the proportion of teachers’ discourse, and the extent to which the classroom conversation is productive, allowing for the characterisation of transitions between non-productive and productive discussions. In the second stage, the qualitative aspects of four different teachers, who have been selected to showcase different ways of using the discourse in the modelling classroom, are explored through the analysis of chronographies, which provide insights into how teacher management of productive conversation is structured within instructional modelling processes. Each of them is a case of the use of dialogue to promote a particular modelling phase or to present the difficulties that can be found in the implementation of a task with modelling potential. This enables a deeper understanding of how productive conversation is facilitated across the phases of the modelling cycle and how dialogue is managed in the classroom.

Teacher orchestration of the modelling classroom conversation

A comparison of the quantitative aspects of the modelling classroom conversation management in the four teachers reveals patterns concerning the transitions between non-productive classroom conversation and productive conversation.

A review of the data reveals that the number of speaking turns in three of the four observed classes (Maria, Chris and Caroline) is comparable (Table 5). In these instances, a substantial proportion of the speech turns, approximately half of which are attributable to the teacher, are characterised by productive conversation, with the teacher's contributions accounting for between 52 and 75% of the total productive speech turns. The analysis demonstrates that modelling is developed through dialogical discursive processes, in which teachers pose questions to students and engage them in collective model construction and refinement processes. Teachers and students alternate speaking turns rapidly and with a high turnover, resulting in a continuous classroom dialogue through brief, punctual, and concise interactions. The rapid succession of interventions indicates a participatory dynamism during the lesson and the potential for ideas to be raised and then quickly reframed. Notably, most of these interventions are productive, indicating a learning-oriented interaction dynamic. Most interventions are focused on meeting instructional objectives and effectively contributing to the co-construction of the model. Furthermore, non-productive classroom conversation, which is also evident in the development of these classes, could be associated with interventions that facilitate the fluency of the instructional modelling process despite not directly contributing to the development of students' models.

Table 5 Dialogue distribution in the implementation of a TLS with modelling potential

	Total turns of speech	Teacher's turns	Teacher's turns in a productive conversation	Teacher's turns in the familiarisation	Teacher's turns in the discussion	Teacher's turns in the consensus
	Total number of communicative acts during the lesson between the teacher and students	Percentage and number of turns that correspond to the teachers	Percentage and number of teachers’ turns corresponding to productive conversation	Percentage and number of teachers’ turns in productive conversation focused on the familiarisation with the phenomenon and the identification of the need for a model.	Percentage and number of teachers’ turns in productive conversation focused on expressing and using ideas to evaluate or review a model.	Percentage and number of teachers turns in productive conversation focused on consensus of ideas to express a final model.
Joseph	104	74% (77)	66% (51)	59% (30)	37% (19)	4% (2)
Chris	534	47% (249)	65% (162)	40% (65)	57% (93)	3% (4)
Maria	569	53% (300)	74% (221)	42% (94)	38% (83)	20% (44)
Caroline	480	55% (265)	52% (137)	58% (80)	42% (57)	0% (0)

---

On the other hand, one of the teachers under examination (Joseph) has a markedly reduced number of speech acts, with a greater focus on the teacher and a comparable prevalence of productive conversational elements in the turns of speech associated with the teacher (Table 5). In this instance, the modelling processes are also developed through dialogic discursive processes, albeit with a greater focus on the teacher, who assumes a more prominent role in the interaction. In the case of Joseph, although productive conversation is present, the reduction in the number of turns of speech and the focus on the teacher may potentially limit the management of the students' modelling processes. This would entail a more autonomous dialogue between the students in order to respond to the task that has been set.

Teaching management of the modelling classroom conversation for each case

An analysis of the speech reveals two conversational patterns that do not align with the differentiation based on the total number of turns of teacher talk. These patterns are discernible only by comparing the teacher's productive speech turns, revealing a distinction between Joseph's, Chris's and, Maria's from Caroline's turns.

A review of these instances of productive conversation reveals that, while some teachers employ the discourse to favour modelling, others appear to struggle with this aspect, often engaging in conversations that inadvertently hinder the modelling process. Furthermore, there are discrepancies in the utilisation of modelling-oriented discourse, with teachers concentrating on facilitating a specific phase of the modelling process rather than the entire cycle.

Based on the previous section's findings, this section presents a case study of teachers who facilitate students' familiarisation, discussion and consensus of ideas to a greater extent, as well as a case study of a teacher who presents difficulties in mediating the modelling process.

(a) Joseph and the persistence of familiarisation. Joseph's lesson is 90 minutes, of which 40 are allocated to classroom conversation on tasks with modelling potential. Joseph sets two tasks with modelling potential (*). Each implemented task follows an incomplete modelling cycle, the nature and duration of which differ significantly. Classroom conversations are productive 43% of the time (Fig. 1).

Fig. 1 Chronography of Joseph's discursive process.

The initial task occupies 28 of the 40 minutes of class time, which includes classroom conversation. This task is dedicated to familiarising the whole students’ group with the studied phenomenon, and the majority of the conversation is not productive.

The second task occupies 12 of the 40 minutes of the class. In this instance, the teacher initiates interaction with the entire class but subsequently engages with the four small student groups organised. In this instance, it is possible to discern the presence of the familiarisation, discussion and consensus phases, although these phases do not occur in all groups. In Groups 1 and 2, complete cycles are observed, with the familiarisation phase predominating and brief moments of discussion and consensus occurring. In contrast, Groups 3 and 4 remain in the familiarisation phase, not engaging in other modelling phases.

Despite this, productive conversation is frequently found to be the predominant form of conversation.

The productive conversation focuses on familiarising students with the phenomenon, while the classroom conversation focuses on familiarising them with laboratory skills and instruments. During this phase, Joseph consistently directs the productive conversation towards eliciting ideas, providing instances in which students feel confident in sharing their thoughts and encouraging them to articulate their ideas as they approach the phenomenon.

The following example illustrates how Joseph provides extensive familiarisation opportunities to students. While carrying out the task “What will be the changes in the reacting substances?” students seek to check how the contact surface of the filament will influence the experiment before making their prediction. This interaction between the teacher and students highlights the process of mediating and supporting a familiarisation with the phenomena that allows students to elicit their model through their predictions.

S1. We want to see what happens when it is less than one gram [of iron filament that has been compacted into a ball].

Joseph. Do you want to try?

S1. Yes [moves a ball of shavings in his hand]

Joseph. You can do it, we still have some solution left so you can check what happens.

Joseph. And how much does that amount to?[points to the ball]

S2. It's what we had available

S1. I don't know, but I want to know what the difference is between the two in the reaction.

Joseph. Okay,

S1. See the reaction when it is less [referring to the iron].

Joseph. But what do you think could happen if now the amount is less?

S2. It's going to have like a…. it's going to be less, it's going to be less, it's going to have a different amount of orange [referring to the colouring of the filament].

Joseph. Ok

S1. It's going to be a different shade with a smaller quantity than with a larger quantity.

Joseph. Let's see what happens, I'm going to get a spoon so you can take it out. [To the copper sulphate solution filament].

(b) Chris and the discussion of the ideas expressed. Chris's lesson lasts 90 minutes, of which 43 minutes correspond to classroom conversation on tasks with modelling potential. Chris sets out four tasks with modelling potential (*). Each task has its own modelling cycle, which differs in nature and duration. Generally, the conversation is productive 56% of the time (Fig. 2).

Fig. 2 Chronography of Chris's discursive process.

The initial task occupies 15 minutes or 35% of the total 43-minute class period. The initial task is exclusively dedicated to the familiarisation phase, which is initially conducted with the entire class and then with a specific smaller group through non-productive classroom discourse. Although the task initiates a dialogue with the entire class and then with a specific small group, the dialogue between the teacher and students is intended to promote familiarisation with the phenomena through a non-productive classroom conversation. This is due to the fact that the instructor's interactions with their students are primarily focused on the delivery of instructions or procedural learning through authoritative dialogue. Moreover, most of the non-productive classroom dialogue is not with the aforementioned small group but rather with other students who interject to request the distribution of materials and responses to general queries.

The second task, which occupied 8 minutes, represented 19% of the total 43-minute class period. The objective of these interactions, which start with the entire class and subsequently break into smaller groups, is to facilitate a transition from a general familiarity with the phenomenon to a more in-depth understanding within the context of smaller groups. This, in turn, enables a productive conversation that includes different ideas of the students.

The third task occupies 3 of the 43 minutes allotted for class time. The objective of this brief task is to facilitate familiarity with the task. However, nearly half of the allotted time is spent responding to general inquiries about the work in progress. This cycle becomes a precursor to the conversations of the next task.

The fourth task occupies 16 minutes or 37% of the 43-minute lesson. This task aims to facilitate a transition from a phase of general familiarisation to a phase of discussion of the ideas in small groups, with one of the groups reaching a consensus phase. In these discussions, the alternation of non-productive and productive conversations allows for managing the classroom climate and resolving general student queries while reaching a common understanding.

In this way, Chris's productive dialogue progresses from a familiarisation phase to a discussion of the students' expressed ideas, in which he promotes using the initial ideas to revise the students’ models. In some instances, this phase entails problematising expressed ideas, elucidating the rationale behind the ideas, utilising them to evaluate the experimental outcomes through evidence or arguments, or contrasting the ideas of students within the same group.

The following example illustrates how Chris provides discussion opportunities to students. While carrying out the task “Represent and/or describe the end state of the reaction,” the teacher checks the small group's response about what happened in the reaction between the ion filament and the copper sulfate solution and their reasoning. This interaction between the teacher and students fosters engagement in a discussion about the ideas largely explored in the familiarisation.

Chris. According to the colour change it has, why does it take on that colour?

Chris. [Waiting time]

S1. What colour change? [Changes the colour of solution and the colour of the filament].

Chris. Of course, because before this, the filament was of this colour, it was grey; but now it is no longer grey, it is reddish.

S1. It rusts

Chris. Yeah, and why does it rust?

Chris. [Waiting time]

S2. It's because it gets hot

S3. Yes, we feel that it heats up.

Chris. Yeah, it could be one of the causes

S1. Is that in order for the metal to become very red, it has to be very hot.

Chris. Yeah, but it's not red-hot. It's only red.

Chris. Now, we would have to think about what is responsible for this change of colour.

S1. It could be because of the sulphate, the sulphate, or copper as one of the factors.

Chris: The copper? Do you attribute it to copper? What is the function of copper?

S1. The sulphate has the same colour as the copper.

Chris. So…

S3. Copper, I mean, like, there's the liquid, right, and the copper sulphate. I mean, they have to take the mineral and when it…

S1. Is that what you wrote down?

S4. It kind of dyes it

S3. Yes, it kind of coats it.

Chris. I mean, if it dyed it maybe it would be blue, and it doesn't look blue. So maybe this apparent change doesn't go hand in hand with the staining, but for example, talking about oxidation could be interesting to discuss.

Also, independent of the tasks, Chris consistently returns to the groups after the conversation to re-examine ideas and assess the potential alterations between the conversation with the teacher and the small group members' conversations. Although he does not revisit all groups after a designated period to assess their ideas, this practice is employed consistently in implementing all tasks.

(c) Maria and the productive conversation towards consensus. Maria's lesson lasts 90 minutes, of which 38 minutes are dedicated to classroom discussion within tasks with modelling potential. Maria proposes three tasks with modelling potential (*). Each task comprises a modelling cycle similar in nature and duration. Overall, the conversation is productive 63% of the time (Fig. 3).

Fig. 3 Chronography of Maria's discursive process.

The three tasks observed in the class are 14, 10 and 14 minutes in duration, respectively. In all three cases, the task initiates with a familiarisation phase involving the entire group, which then progresses to the small groups and constitutes the majority of the time allotted to modelling instruction. Subsequently, the discussion phase occurs within the same groups. Once the ideas have been discussed, the teacher initiates a dialogue with the entire group to facilitate the consensus phase, which is the phase that devotes the least amount of time. Consequently, each small group is engaged in the entirety of the modelling cycle for each task through interactions with the teacher, which alternate between productive and non-productive classroom conversations.

In Maria's productive conversation, there is a discourse that facilitates students' engagement in modelling processes. This is evidenced by the presence of an instructional pattern that favours modelling, whereby time is explicitly allocated to familiarisation, discussion, and a consensus on the ideas proposed not only in the small groups but also on the synthesis and reorganisation of these ideas, taking into account the contrasting perspectives presented in the whole group discussion. In instances of consensus in the productive conversation, Maria focused on promoting the compilation of the students' ideas to express a final model. She also requested that the final representations be reviewed among the students in different small groups. While these instances aim to arrive at a collective understanding, the teacher occasionally provides the initial stimulus for formulating a final answer to the task assignment. The following example illustrates how Maria provides discussion opportunities to students. While carrying out the task “Represent the final state after doing the experiment (micro),” the teacher asks students to elaborate on their new representations to check for a consensus on the drawings. This interaction between the teacher and students encourages dialogue to arrive at a common answer for the task.

Maria. Right, question then for you, was the drawing you did now the same drawing you had done at the beginning?

Several students. No

Maria. No, What changes did you have?

S1. Many

Maria. Many Like what?

Maria. [Waiting time]

S2. It used to be yellow

S2. The copper

Maria. The copper, what happened to the copper?

S3. What's left over from the filament

Maria. Yes, we had said that we had to draw what was left over from the filament, right?

S4. What reacted as well

Maria. At the beginning you also drew more cupric sulphate molecules, didn't you? and we had to draw more?

Several students. No

Maria. No, didn't we?We had to draw the same amount, but we had to see how it was arranged in the glass.

(d) Caroline and the conversation that hinders the construction of the model. Caroline's lesson lasts 60 minutes, of which 36 are dedicated to classroom conversation on tasks with modelling potential. Caroline organises her lesson, including six tasks with modelling potential. Each task is associated with a distinct modelling cycle, which varies in terms of its intrinsic nature and duration. Generally, the conversation is productive 35% of the time (Fig. 4).

Fig. 4 Chronography of Caroline's discursive process.

The first task accounts for 4 of the 36 minutes of the class. The initial conversation is primarily focused on the familiarisation phase. Despite the limited duration of the cycle, the teacher engages in dialogue with three small groups with the aim of familiarising them with the phenomenon. However, two-thirds of the dialogue between the teacher and the students for familiarisation within the groups is non-productive classroom conversation. This is because the teacher's interactions with her students focus on giving instructions and raising expectations about the subsequent discussion of the results rather than the conversation about their ideas for the phenomenon.

The second and third tasks, respectively, account for 4 and 11 minutes of the 36 minutes of class time. The dialogue in these tasks is primarily characterised by non-productive classroom conversations, with a significant proportion of the time dedicated to setting expectations for subsequent discussions on ideas, directing attention to ongoing work, and providing general instructions. In both tasks, the initial focus is on familiarisation, beginning with a dialogue among the whole group and progressing to small group conversations. However, productive conversation is not a constant feature of the dialogue, as it alternates with non-productive classroom conversation, which makes maintaining the continuity of dialogue challenging. Although the third task includes an instance for the discussion phase, most of the dialogue is non-productive, seeking to postpone the discussion until later.

In the fourth task, which occupied only 2 of the 36 minutes of the class, the conversation again turned to the familiarisation phase, with most of the conversation occurring as a non-productive classroom conversation involving the entire group. Subsequently, a brief dialogue is initiated with a small group.

The dialogue concerning the non-productive classroom conversation is aligned with the teacher's instructions regarding subsequent instances in which the students' ideas can be discussed.

In the fifth task, which occupies 4 of the 36 minutes of the lesson, the focus on the familiarisation phase is maintained; however, it can be observed that this task comprises a dialogue in which the teacher engages in conversation with a small group. In this dialogue, non-productive classroom conversation accounts for approximately 80% of the dialogue time focused on answering queries, providing general instructions or reminding students that the subsequent discussion is contingent upon their enquiry about the phenomenon.

The sixth task occupies 12 of the 36 minutes of the lesson. In this case, the teacher initiates a dialogue with the students in the whole group, progresses to a discussion with a small group and then reverts to the whole group. The initial dialogue with the whole group is oriented towards familiarisation. The initial familiarisation phase is maintained upon resuming the dialogue with the smaller group. However, upon transitioning to the discussion phase, Caroline identified a deficiency in the student's comprehension.

This prompted her to have a non-productive classroom conversation centred on providing general instructions and requesting that the students switch to different lab materials. In this non-productive classroom conversation, the teacher resumes the dialogue with the whole group, continuing to revisit the initial familiarisation phase before initiating a discussion. Approximately 30% of the discussion is devoted to non-productive classroom conversation, which pertains to an authoritative dialogue in which the teacher reviews the scientific ideas the students are expected to grasp to continue the conversation. Given the above, productive and non-productive conversations alternate.

In contrast to the other teachers, Caroline's productive dialogue with her students does not focus on any of the phases or the completion of an entire modelling cycle. In addition, maintaining a consistent approach to opening the conversation and managing the students' modelling process is challenging. This is evident in the alternating dialogue between non-productive classroom conversation and productive conversation that she has with the students, both in whole-group settings and in small groups.

The following example illustrates how productive dialogue is embedded within classroom conversation. While carrying out the task, “What do you think the final state of the reaction will be like?” Caroline checks the small group's prediction about the possible outcome of the reaction between the ion filament and the copper sulphate solution.

This interaction between the teacher and the students shows the teacher's difficulties in creating a productive conversation as she misses opportunities to engage with the students' ideas.

Caroline: What was it that you expected to observe?

S1. I, I thought it was going to dissolve the…

S2. How was that going to happen?

Caroline. Like a dissolution?

S1. Right, as if it was a dissolution that…

Caroline. Yeah, but this is, is it a…?

S2. Oxidation

Caroline. …Chemical reaction

Caroline. A REDOX reaction, even though there wasn't going to be any…

S1. There wasn't going to be any…

Caroline. There wasn't going to be any metal left, and there was going to be a dissolution.

Caroline. And you wrote that down in the predictions?

S1. Yes

Caroline. Yeah, the idea is that you write it down.

S1. I am…

Caroline. But are you reaching a consensus as a group?

S2. Eh

Caroline. Or is everyone making observations?

S2. Eh, more or less, but we are still here talking.

Caroline. Yeah, and then you talk about it

S2. Of course

The management of the productive modelling conversation

The analysis undertaken allows us to explore the different ways in which the productive conversation of four chemistry teachers who implemented the same teaching and learning sequence they had co-designed was managed. The results clearly show the personal adjustments made by each teacher in terms of both the number of tasks implemented and the classroom management, which fluctuates between working with the whole group and working in small groups.

From a discursive perspective, the class occurs in dialogical settings, characterised by a continuous dialogue between the teacher and the students, both in whole class and small groups. Nevertheless, significant discrepancies in the orchestration of this dialogue influence the modelling process differently.

Teachers facilitate the dialogue by alternating between non-productive classroom conversation and productive conversation, monitoring students' progress in transitioning between the modelling phases and engaging in the students' conversation to advance the ideas under discussion. As the class progresses, the teacher's role and how they organise the students undergo a series of changes with the objective of facilitating the processes of familiarisation, discussion and consensus of ideas in the modelling conversation. In the course of our study, we came to recognise that the manner in which the teacher oversees the productive conversation and makes the students engage in the various phases of modelling can either facilitate or pose an obstacle to the modelling processes. This has enabled us to identify the key aspects of classroom management in the context of modelling conversations, which we discuss in the following section.

Discussion

The analysis of the productive modelling conversation of four chemistry teachers has enabled the characterisation of different ways teachers’ classroom discourse promotes modelling processes in the chemistry classroom. The main findings of this study focus on four dimensions of the conversation in the classroom that occur when different teachers promote students' modelling practices and construction of models:

(a) The role of the teacher in the directionality of classroom conversations, examined through the number and distribution of teacher and students' turns of speech.

(b) The presence of non-productive discourse in modelling processes, interpreted by the organisation of the class and the typology of classroom interactions (categories b and c).

(c) The identification of specific moments that initiate shifts in the teacher's discursive trajectory, focusing on transitions between general classroom discourse and productive modelling conversation (category b).

(d) The construction of the model through complete or incomplete instructional modelling cycles, by conceiving the modelling process as a whole composed of successive and interconnected learning cycles (category d).

Regarding the directionality of the conversation, the analysis of the turns of speech in the four teachers’ lessons reveals an equal distribution between the teacher and the students and a considerable number of speech turns throughout the lesson (Table 5). An example to illustrate this fact is the case of Maria's conversations, in which the consensus phase involves 96 turns of speech that occur over 6 minutes. Of those turns of speech, 44 correspond to Maria, which evidences that she is constantly mediating the conversation (Table 5 and Fig. 3). These conversations have a high degree of turnover, where the teacher makes almost half of the turns. Colley and Windschitl (2016) argue that as more turnover is observed, it is more likely that teachers are using students’ contributions to promote modelling performance. Therefore, the teachers' high number of turns of speech (Table 5) is an indicator of a rigorous conversation, but also a way of understanding that the conversation and its direction depend on what teachers understand of their students' reasoning and how they engage with these understandings to promote learning.

When reviewing the presence of discourse that cannot be categorised as productive, we need to go beyond the frequency of this dialogue to look at its function (for instance, as shown in the shared quotes and reported along each teacher's case); this dialogue is not only quick but intensely intentional. In this sense, Maria's frequent turns of speech serve as a mediation that would allow the expression of a model and change the directionality of a conversation while she is responding to her students' ideas and learning needs. This would mean that teachers' frequent interventions are indicative of their role as active mediators, who are highly involved in the conversation and the directionality of the dialogue during class. We can relate this fact to the expected role of teachers as truly activators of learning instead of mere facilitators or guiders.

Many of these discourse patterns comprise examples of teachers' utilisation of dialogic strategies to render students' ideas visible and evaluate their responses, thereby determining how to maintain and manage the conversation. Such as when Maria asks, “What happened to the copper?” or Chris says “Yeah, but it's not red-hot. It's only red”. In this sense, the teachers are constantly engaged in the process of determining how to incorporate the students' ideas into the discourse in a manner that mediates the construction of ideas that align with the school's scientific model (Lehesvuori et al., 2013; Colley and Windschitl, 2016). In that regard, the incorporation of students' ideas refers to how teachers establish the trajectory and continuity of the discourse. While one might think that the teachers are only monitoring the reasoning in terms of scientific correctness, they are also thinking of how they can guide the conversation in their role as activators (Hamnell-Pamment, 2023) in a way that would enable them to determine how to continue the conversation responsively.

Consequently, it could be expected that responsiveness in the instructional modelling classroom would result in more ideas from the students that are anchored to the phenomena and the continuation in the use of those ideas across the implementation of the tasks; however, they are not the only kind of responses we could expect from teachers (Gray et al., 2022). This responsiveness can be seen in Joseph's lesson when students ask for new opportunities for familiarisation; not only does he promote an extra step in the experiment, but he asks questions about the student's ideas that go beyond the demand of the task: “But what do you think could happen if now the amount is less”.

Consequently, we can associate the role of activator with responsiveness in the conversation, where exploring the student's ideas would lead the teacher to decide the direction of the conversation, for which a high number of speech turns would be needed. Thus, according to our results, the directionality of the conversation in modelling science lessons is primarily influenced by the teachers’ discourse and their frequent requests for ideas and engaging with students' ideas throughout the lesson.

The productive conversations between teachers and students are contingent upon the teacher's ability to manage instances of model construction in which they are not privy to the entirety of the dialogue held by the students. This can be illustrated by the case of Joseph, Chris and Maria, where we observe that in alternating productive and non-productive conversations, they demonstrate responsiveness to what they hear in their students' dialogue before intervening (Fig. 1–3). In contrast, in Caroline's case, the alternations are characterised by a higher presence of non-productive conversations and shorter conversations with small groups, which create a constant switch that impacts the continuation of productive dialogue (Fig. 4). The discontinuities in the conversations Caroline maintains with her students stem from her instructional decisions about the need to remind students about classroom expectations or a lack of responsiveness, such as when she does not use her students' ideas. Those alternations between productive and non-productive classroom conversations occur intrinsically in every teacher's discourse. These moments of non-productive classroom conversation relate to all the interactions where the focus is not on constructing the model through dialogic/interactive conversation, so it references non-interactive or authoritative dialogue. Such as when Caroline says, “Yeah, the idea is that you write it down”. Although these decisions align with the communicative approach of the teaching process that teachers rely on (Lehesvuori et al., 2013; Hamnell-Pamment, 2023), this may lead them to repeatedly engage with the same ideas during the modelling process. In that sense, the teacher's responsiveness to the learning opportunities students need could impact the flow of the dialogue and create a switch between productive and non-productive class conversations.

This transition is essential for the continuation of the productive conversation. The teachers may need to abandon a productive discussion to respond to students' learning necessities, making way for this alternation, which occurs naturally and would make it sustainable. Nevertheless, the alternation between productive and non-productive conversation indicates the presence of an ongoing dialogue whose purpose is to facilitate the monitoring of students' progress and the mediation of the joint construction of the model.

Concerning what triggers the teachers’ discursive dialogue trajectory, it needs to be acknowledged that the above-mentioned dimensions of discourse management relate to instances where teachers constantly identify issues in their students' interventions. These issues prompt teachers to believe it is necessary to intervene in students' interactions. Furthermore, teachers decide whether their intervention in students' conversations should consist of a productive or non-productive classroom conversation. As a result, Joseph, Chris, Maria and Caroline engaged in instructional modelling processes that not only maintain and manage an interactive dialogue, allowing students to construct models through productive conversation, but also require the creation of spaces in which the dialogue is non-productive (Fig. 1–3).

The latter relates to the idea that every non-productive classroom conversation has the potential to be considered an authoritative dialogue, even though the teachers are responding to the learning needs of students. Responsive discourse management in the instructional modelling classroom leads to different approaches to the conversation, which go beyond the common aspects we have discussed so far, characterised by frequent interventions by teachers and alternation between productive and non-productive conversation. These responses are driven by the perception of learning needs, which are prompted by specific situations or interventions. As a result, a key aspect of the teacher's role in the conversation would be the teacher's capacity to maintain a productive conversation at the right moment (Buty and Mortimer, 2008). The latter would enable teachers to recognise the ideas or skills required by students to facilitate meaningful learning in science (Lehesvuori et al., 2013). In this way, teachers can continue the productive conversation, understood as interactive dialogue, through non-interactive or authoritative dialogues, using students' ideas in relation to the school scientific model to determine the trajectory of the discourse (Chin, 2007; Lehesvuori et al., 2013). While we can identify instances where student answers prompt teachers to initiate non-interactive dialogue, we cannot ascertain the specific triggers within the scope of this study. What we know is that these conversations are structured in a way that responds to the provision of learning opportunities, which are essential for students to engage in a dialogue that allows them to construct a school science model. This means that in order to facilitate the modelling process, engaging in non-interactive or authoritative dialogical conversations would be necessary. Both conversation types can be observed in teachers who promote modelling practice in the classroom; however, the pivotal factor might be identifying the opportune moments for productive discourse. In this sense, we assume that a higher percentage of productive conversation does not seem to be related to an enhanced modelling process. Instead, we are inclined to believe that the pivotal factor is the teacher's capacity to discern the optimal opportune moments for productive discourse. This is consistent with previous reports in traditional classrooms and modelling-oriented classrooms (Chin, 2007; Buty and Mortimer, 2008).

The trajectory and focus of the conversations go beyond the conversational style of the teachers, as they represent the ways in which the teachers respond to the learning needs of their students. In that respect, in a classroom where teachers want to promote modelling, one could expect the teachers to promote complete cycles in which teachers give the space to actively participate in every phase of the modelling cycle (Garrido, 2016). However, we can observe that, most of the time, teachers' instruction follows incomplete cycles where the implemented phases correspond to familiarisation and discussion (Vergara, 2022). In Maria's case, we can observe complete cycles for every task (Fig. 3), while in the cases of Joseph (Fig. 1), Chris (Fig. 2) and Caroline (Fig. 4), the cycles are often incomplete, whether they focus on the familiarisation within a task, or they end the conversations in the discussion phase. In this sense, the ideas explored, discussed and agreed upon depend on the teachers' choices for the continuation of the students' construction of school science models.

Although we might think that a task with an incomplete modelling cycle might limit the construction of the model, the ideas expressed in the familiarisation phase of one task may be part of the discussion of another task, and this discussion may be part of how students reach consensus on an entirely different task. Thus, a task with an incomplete cycle, where the conversations are focused on a particular phase of the modelling process, contributes to progress in constructing a model. In this manner, modelling cycles are not independent but make sense when analysed as a whole, where ideas and discussions from one modelling cycle feed into the next. In this way, a cycle does not have to be completed, and modelling processes do not occur in a single cycle that progresses wholly and linearly (Vergara et al., 2025). However, processes are opened and closed in a somewhat iterative logic. This is not a minor issue because it concerns how we conceive the modelling cycles as teaching guidelines that can be applied flexibly as long as the teacher understands the nature and purpose of each modelling phase. Thus, the construction of the model through students' ideas could not be forced; hence, progressing from one phase of the modelling cycle to the next when students are not adequately prepared could hinder students' progress.

Ultimately, we encounter highly distinctive dialogic processes that permeate modelling opportunities. These range from how chemistry teachers implement a co-designed teaching and learning sequence (Table 5), their ways to manage discourse within tasks, and their discourse alternations to facilitate transitions between modelling phases (Fig. 1–4). It is important to ascertain whether teachers engage in these processes with a conscious and intentional approach, adapting their instruction based on student feedback, or whether their responses are more reactive, driven by uncertainty or a lack of alternative strategies (examples of dialogue for every teacher). This ambiguity highlights the complexity of teaching practices in fostering a social construction of models. All in all, the findings discussed here signal the main characteristics of what could be referred to as “modelling conversation”: the type of conversation in which teachers engage when using modelling-based teaching and learning sequences and tasks in order to promote students' modelling practices and the construction of models.

Limitations

We acknowledged that, due to the relatively limited number of teachers included in this research, it is not feasible to generalise the findings regarding managing productive conversation in modelling classrooms. Nevertheless, this contributes significantly to understanding the dialogical processes between science teachers and their students and what enables the social construction of scientific ideas in modelling-oriented discussions. The reporting of the ways in which teachers manage and scaffold modelling processes in students could serve as a foundation for the study of communicative acts in which the expression, revision and adjustment of a school science model is sought. It also opens the possibility of studying how teachers' discourse impacts students' discourse in the instructional modelling classroom. Additionally, the study of teachers' speech turns in the modelling classroom is an upcoming area that requires research in order to understand how students' representations are refined through teachers' discourse.

While dialogue in modelling contexts has yet to be investigated at the micro level with talk moves, understanding how science teachers focus the conversation at the level of communicative events in the phases of the modelling cycle during a lesson can provide insights into how teachers conceptualise the dialogic process that students must undertake in order to construct a scientific school model.

On the other hand, in this study, we use “non-productive” to refer to speech turns that do not align with productive classroom conversation. Nevertheless, this level does not mean they lack value or cannot support learning. We recognise that labelling discourse as “non-productive” may lead to misinterpretations, as classroom conversations naturally include both productive (interactive/dialogical) and other types of interactions. Rather than defining them by what they lack, we acknowledge the need for a typology that captures their own characteristics.

Conclusions and implications

This study has illuminated key characteristics of classroom discourse that support students' modelling practices in chemistry education. Through the analysis of four teachers' conversations in modelling-based instruction, four interrelated dimensions were identified that shape discourse: the teacher's role in guiding conversations, the presence and function of non-productive dialogue, the discursive triggers that shift dialogue trajectories, and the idiosyncratic use of instructional modelling cycles. The findings of this study contribute to a more profound understanding of the manner in which discourse mediates model construction and learning in science classrooms.

An insight derived from this analysis is teachers' central role in orchestrating the directionality of conversations. The high frequency and alternation of teacher and student speech turns highlight how teachers use dialogic strategies to evaluate student ideas and actively steer the discourse toward instructional modelling. Teacher discourse is revealed as a dynamic mechanism of responsivity, whereby teachers act as activators who shape and reshape the conversation in response to students' reasoning. This responsiveness is found to be essential for sustaining instructional modelling conversations and fostering student engagement in model expression, evaluation and refinement.

Furthermore, the analysis challenges the binary categorisation of productive versus non-productive discourse. Non-productive episodes, frequently designated as non-interactive or authoritative, have been shown to fulfil significant instructional functions, such as re-establishing classroom expectations or clarifying conceptual confusion. Rather than disrupting the learning process, these interjections can provide the necessary scaffolding to facilitate the continuation of productive dialogue. It is important for teachers to recognise when such interventions are necessary to maintain the momentum of the modelling process.

Concerning modelling cycles, it is evident from the findings that complete cycles are not always carried out within a single task or lesson. While Maria's enactments exhibit complete modelling cycles, other teachers implement fragmented yet meaningful segments of the cycle. This finding suggests that modelling should be regarded as an iterative process that extends across tasks and time rather than a strictly linear or self-contained sequence. Incomplete cycles can still contribute significantly to model construction when teachers and students can recontextualise ideas across different phases and tasks. Our study in a normal classroom setting reinforces this idea already found in teacher education and expert teachers’ contexts (Garrido and Couso, 2024; Vergara et al., 2025).

The notion of “modelling conversation” that emerges from this study captures the nuanced and flexible dialogic processes involved in modelling-based instruction. Teachers' ability to move between interactive and authoritative acts, respond to student thinking in real-time, and strategically manage the modelling cycle are all decisive in creating learning environments where scientific models can be co-constructed. Even though our focus is on the implementation of a design with modelling potential (Cortés-Morales and Marzábal, 2025), the teacher's role in the conversations leads us to believe that good design is not enough; discourse makes a difference, as it is ultimately about what and how teachers do to promote students modelling performance. Future research should further explore the intentionality behind teachers' discourse moves and the extent to which these practices are developed, supported, and sustained through professional learning and collaborative design.

The latter leads us to believe that sustained implementation of modelling-based instruction requires a dual emphasis on teachers' discursive practices and professional development that attends to the discursive dimensions in teaching. For research, the findings of this study underscore the importance of research about how teachers recognise and enact a productive conversation and the need for a framework of discursive moves oriented towards modelling instruction. A deeper understanding of the relationship between discourse, design and teacher discursive decisions would be essential for advancing the theoretical and practical bases of modelling-based instruction. For teaching, this means the need to cultivate discursive awareness and a strategic use of discourse. Professional learning opportunities should provide spaces for teachers to analyse classroom discourse, reflect on the interactions with the students and refine their conversational strategies to support modelling.

Author contributions

A. Cortés-Morales: Conceptualisation, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, supervision, validation, visualisation, writing – original draft, writing – review & editing. A. Marzabal: Conceptualisation, formal analysis, investigation, validation, visualisation, writing – review & editing. D. Couso: conceptualisation, formal analysis, investigation, validation, visualisation, writing – review & editing.

Data availability

The data for the article cannot be made available due to ethical confidentiality requirements.

Conflicts of interest

There are no conflicts to declare.

Appendix

Appendix 1 Coding of an extract according to Table 4 codes and keys.

Teacher: Maria

Task: Represent the final state after doing the experiment

---

Degree of openness	Science practice	Context	Modelling potential
Open	Verification	Scientific	Yes

---

Modelling phase	Actor	Organisation	Time	Dialogue	Typology
					CI^1	CT^2	AJA^3	M^4	TRM^5	NPC^6
F	Teacher	Whole group	25:27	Now, let's do the second drawing, it says ‘Represent (draw again) the final state after doing the experiment’ So, now let's go do another drawing; let's see if it is indeed what we thought was going to happen, thinking about the reactants, sorry, thinking about the products. Now, incorporate the filament and see what happens.	X
F	Teacher	Whole group	25:55	[Give students time to think]			X
F	Teacher	Whole group	26:09	Now, to do the drawing, pay attention here to give you a fact. This is the cupric sulphate that we have been using (shows the volumetric flask). It is the same one we always use. Or did we change it, Jeremy?(lab assistant, who says no with his head)						X
F	Teacher	Whole group	26:40	So, if it's the same, do we have to add to the number of balls we draw to represent it, or do we have to decrease it?			X
F	Student 1	Whole group	26:50	No
F	Teacher	Whole group	26:51	Or did we draw it the same?			X	X
F	Various Students	Whole group	26:52	The same
F	Teacher	Whole group	26:54	It's drawn the same because the only thing that changed was the volume, not the concentration.	X
F	Teacher	Whole group	27:03	Yeah, so when you draw your second representation now, we have to consider then that the sulphate is the same as the one we have been working with in previous experiments. The only thing that has been changing is the volume.	X
F	Teacher	Group 1	27:21	So, your glass [speaking to group 1] with sulphate, does it cover the top of the glass as you drew it?			X
F	Student 2	Group 1	27:27	No
F	Teacher	Group 1	27:28	How far does it cover?		X
F	Student 2	Group 1	27:29	No, to less than half

1. CI – Clarity and intelligibility: Refers that there should be healthy and elaborated intellectual exchanges among peer community by talk instances of probing, requesting clarification or revoicing responses to eliminate clarity issues.

2. CT – Critiques in the talk: Indicates that alternative points of view should be considered and rigorously negotiated during classroom discourse, by pointing out and guiding students in reviewing the ideas and arguments made explicit.

3. AJA – Accountability-Justification-Authority: Stands for that student-led or teacher-led talks should incorporate a version of accountability or epistemic responsibility by motivating and promoting how students are considering evidence to defend their position, while evaluating others' ideas.

4. M – Intense discursively- oriented meta-cognitive activity: Refers to all interactions that should be synchronised or paralysed during discussions for meaningful learning by promoting students to notice the changes in the discourse and the featured ideas and how they progress.

5. TRM – Teacher as the discursive role model: Stands for science teacher as a discursive role model in scaffolding students’ scientific practices by modelling and rehearsing procedures in ways that students act in multivariable thinking.

6. NPC – No productive conversations: Refers to turns of speech that do not comply with the other categories.

Acknowledgements

This research was funded by the grant programme of the Chilean National Research and Development Agency (Agencia Nacional de Investigación y Desarrollo: ANID) under Grant Doctorado Nacional 2021 number 21210524, as well as the Ministerio de Ciencia e Innovación of Spain (PID2022-138166NB-C22b) and carried out within the SGR ACELEC research group, ref. 2021 SGR 00647w.

References

1. Alexander R., (2018), Towards dialogic teaching: Rethinking classroom talk, Dialogos.
2. Buty C. and Mortimer E. F., (2008), Dialogic/authoritative discourse and modelling in a high school teaching sequence on optics, Int. J. Sci. Educ., 30(12), 1635–1660 DOI:10.1080/09500690701466280.
3. Chin C., (2007), Teacher questioning in science classrooms: approaches that stimulate productive thinking, J. Res. Sci. Teach., 44(6), 815–843 DOI:10.1002/tea.20171.
4. Colley C. and Windschitl M., (2016), Rigor in elementary science students’ discourse: the role of responsiveness and supportive conditions for talk, Sci. Educ., 100(6), 1009–1038 DOI:10.1002/sce.21243.
5. Cortés-Morales A. and Marzábal A., (2025), The design of science teaching and learning tasks with modelling potential, Stud. Sci. Educ., 1–29 DOI:10.1080/03057267.2025.2493990.
6. Garrido A. and Couso D., (2024), The IPM cycle: an instructional tool for promoting students’ engagement in modelling practices and construction of models, J. Res. Sci. Teach., 62(2), 391–425 DOI:10.1002/tea.21979.
7. Garrido Espeja A., (2016), Modelització i models en la formació inicial de mestres de primària des de la perspectiva de la pràctica cienttfica, TDX (Tesis Doctorals en Xarxa), https://www.tdx.cat/handle/10803/399837.
8. Gilbert J. and Justi R., (2016), Modelling-based teaching in science education, Springer.
9. Gobert J. D. and Buckley C., (2000), Introduction to model-based teaching and learning in science education, Int. J. Sci. Educ., 22(9), 891–894 DOI:10.1080/095006900416839.
10. Gouvea J. and Passmore C., (2017), ‘Models of’ versus ‘Models for’, Sci. Educ., 26, 49–63 DOI:10.1007/s11191-017-9884-4.
11. Gray R. and Rogan-Klyve A., (2018), Talking modelling: examining secondary science teachers’ modelling-related talk during a model-based inquiry unit, Int. J. Sci. Educ., 40(11), 1345–1366 DOI:10.1080/09500693.2018.1479547.
12. Gray R., Rogan-Klyve A. and Canipe M. M., (2022), Investigating the impact of eliciting and being responsive to students’ initial ideas on productive disciplinary engagement across a unit, Sci. Educ., 106(2), 312–334 DOI:10.1002/sce.21701.
13. Guy-Gaytán C., Gouvea J. S., Griesemer C. and Passmore C., (2019), Tensions between learning models and engaging in modeling. Exploring implications for science classrooms, Sci. Educ., 28(8), 843–864 DOI:10.1007/s11191-019-00064-y.
14. Hamnell-Pamment Y., (2023), Making sense of chemical equilibrium: productive teacher-student dialogues as a balancing act between sensemaking and managing tension, Chem. Educ. Res. Pract., 25(1), 171–192 10.1039/d3rp00249g.
15. Hennessy S., Rojas-Drummond S., Higham R., Márquez A. M., Maine F., Ríos R. M., García-Carrión R., Torreblanca O. and Barrera M. J., (2016), Developing a coding scheme for analysing classroom dialogue across educational contexts, Learn. Cult. Soc. Interact., 9, 16–44 DOI:10.1016/j.lcsi.2015.12.001.
16. Johnstone B., (2000), Reading and listening: discourse analysis, In Qualitative Methods in Sociolinguistics, Oxford University Press, pp. 103–129.
17. Kim M., (2020), Teacher scaffolding strategies to transform whole-classroom talk into collective inquiry in elementary science classrooms, Alberta J. Educ. Res., 66(3), 290–306 DOI:10.55016/ojs/ajer.v66i3.56957.
18. Lehesvuori S., Ramnarain U. and Viiri J., (2018), Challenging transmission modes of teaching in science classrooms: enhancing learner-centredness through dialogicity, Res. Sci. Educ., 48(5), 1049–1069 DOI:10.1007/s11165-016-9598-7.
19. Lehesvuori S., Viiri J., Rasku-Puttonen H., Moate J. and Helaakoski J., (2013), Visualizing communication structures in science classrooms: tracing cumulativity in teacher-led whole class discussions, J. Res. Sci. Teach., 50(8), 912–939 DOI:10.1002/tea.21100.
20. Marzábal A., Merino C., Soto M. and Cortés-Morales A., (2024), Modeling-based science education, in Marzabal A. and Merino C. (ed.), Rethinking Science Education in Latin-America, Contemporary Trends and Issues in Science Education, Cham: Springer, vol. 59.
21. Merino C. and Izquierdo M., (2011), Aportes a la modelización según el cambio químico, Educ. Quím., 22(3), 212–223 DOI:10.1016/s0187-893x(18)30137-x.
22. Michaels S. and O’Connor C., (2015), Conceptualizing talk moves as tools: professional development approaches for academically productive discussions, Socializing Intelligence Through Academic Talk and Dialogue, American Educational Research Association, pp. 333–347.
23. Michaels S., O’Connor M. C., Sohmer R. and Resnick L., (2009), Guided construction of knowledge in the classroom: teacher, talk, tasks and tools, in Transformation of Knowledge through Classroom Interaction, Routledge, pp. 105–129.
24. Mortimer E. F. and Scott P., (2003), Meaning making in secondary science classrooms, Open University Press DOI:10.1002/sce.20103.
25. Oliva J. M., (2019), Distintas acepciones para la idea de modelización en la enseñanza de las ciencias, Enseñ. Cienc. Rev. Investig. Exp. Didáct., 37(2), 5–24 DOI:10.5565/rev/ensciencias.2648.
26. Palincsar A., (1998), Social constructivist perspectives on teaching and learning, Annu. Rev. Psychol., 49(1), 345–375 DOI:10.1146/annurev.psych.49.1.345.
27. Park J., Michaels S., Affolter R. and O’Connor C., (2017), Expanding the conversation: traditions, research, and practice supporting academically productive classroom talk, Oxford Research Encyclopedia of Education.
28. Patton M. Q., (2015), Sampling, qualitative (purposeful), The Blackwell Encyclopedia of Sociology, Blackwell Publishing Ltd, pp. 4006–4007 DOI:10.1002/9781405165518.wbeoss012.pub2.
29. Shi F., Wang L., Liu X. and Chiu M. H., (2021), Development and validation of an observation protocol for measuring science teachers’ modeling-based teaching performance, J. Res. Sci. Teach., 58(9), 1359–1388 DOI:10.1002/tea.21712.
30. Soysal Y., (2019), Indicators of productive classroom talk and supporting discourse moves: a systematic review for effective science teaching, Acad. J. Educ. Sci., 3(2), 114–137 DOI:10.31805/acjes.642246.
31. Téllez-Acosta M. E., McDonald S. and Acher A., (2023), Sociocultural ways of conceptualizing professional vision to support and study novice science teachers’ learning, Int. J. Sci. Educ., 1–21 DOI:10.1080/09500693.2023.2290602.
32. Vergara C., (2022), Análisis del discurso docente en actividades de modelización sobre fuerzas y movimiento, TDX (Tesis Doctorals en Xarxa), https://www.tdx.cat/handle/10803/688282.
33. Vergara C., López-Simó V. and Couso D., (2025), El ciclo de modelización a través de la interacción oral en talleres experimentales sobre fuerza y movimiento, Rev. Eureka Enseñ. Divulg. Cienc., in press.
34. Windschitl M. A., Thompson J. and Braaten M. L., (2018), Ambitious science teaching, Harvard Education Press.
35. Windschitl M., Thompson J., Braaten M. and Stroupe D., (2012), Proposing a core set of instructional practices and tools for teachers of science, Sci. Educ., 96(5), 878–903 DOI:10.1002/sce.21027.
36. Woodsong C., Mack N., MacQueen K., Guest G. and Namey E., (2005), Qualitative research methods: a data collector's field guide, Family Health International.
37. Yin R. K., (2003), Case study research. Design and methods, SAGE Publications.

---