*Article* **Reasoning on Controversial Science Issues in Science Education and Science Communication**

**Anna Beniermann \* , Laurens Mecklenburg and Annette Upmeier zu Belzen**

Biology Education, Humboldt-Universität zu Berlin, 10099 Berlin, Germany; laurens.mecklenburg@web.de (L.M.); annette.upmeier@biologie.hu-berlin.de (A.U.z.B.)

**\*** Correspondence: anna.beniermann@hu-berlin.de

**Abstract:** The ability to make evidence-based decisions, and hence to reason on questions concerning scientific and societal aspects, is a crucial goal in science education and science communication. However, science denial poses a constant challenge for society and education. *Controversial science issues* (CSI) encompass scientific knowledge rejected by the public as well as *socioscientific issues*, i.e., societal issues grounded in science that are frequently applied to science education. Generating evidence-based justifications for claims is central in scientific and informal reasoning. This study aims to describe attitudes and their justifications within the argumentations of a random online sample (*N* = 398) when reasoning informally on selected CSI. Following a deductive-inductive approach and qualitative content analysis of written open-ended answers, we identified five types of justifications based on a fine-grained category system. The results suggest a topic-specificity of justifications referring to specific scientific data, while justifications appealing to authorities tend to be common across topics. Subjective, and therefore normative, justifications were slightly related to conspiracy ideation and a general rejection of the scientific consensus. The category system could be applied to other CSI topics to help clarify the relation between scientific and informal reasoning in science education and communication.

**Keywords:** argumentation; reasoning; justifications; socioscientific issues; societally denied science; controversial science issues; science communication; science education

### **1. Introduction**

The OECD Learning Compass 2030 [1] highlights the rapid changes confronting our society and, consequently, the importance of adaptive education in formal and informal learning environments. It emphasizes the need to think and act responsibly "towards collective well-being" [1] based on knowledge, attitudes, values, and skills (including reasoning and critical thinking) as a 21st-century goal [1]. In contrast to this goal, science denial poses a constant or even growing challenge for society [2] and science education [3]. Informed citizens should be able to make evidence-based decisions on questions concerning scientific and societal aspects, e.g., health and environmental issues [4].

The inevitable connections between science and society in science education are bundled under the term *socioscientific issues* (SSI), defined as "societal issues with conceptual or technological ties to science" [5]. SSI are scientific topics that are often discussed controversially by the public [6]. They are well-acknowledged as contexts for science learning [7,8], as the SSI approach integrates scientific, sociological, and ethical content to foster reasoning on complex questions [9]. For example, the current COVID-19 pandemic illustrates the rise in controversy between society and science and, moreover, in doubt about scientific findings [10].

While the scientific foundation of some SSI is mostly accepted by the public (e.g., knowledge about stem cells), controversy may arise with the ethical dilemmas of its application (e.g., stem cell research for medical purposes) [11]. Other SSI are based on societally controversial science that may even be rejected by parts of the public (e.g., anthropogenic causes

**Citation:** Beniermann, A.; Mecklenburg, L.; Upmeier zu Belzen, A. Reasoning on Controversial Science Issues in Science Education and Science Communication. *Educ. Sci.* **2021**, *11*, 522. https://doi.org/ 10.3390/educsci11090522

Academic Editors: Moritz Krell, Andreas Vorholzer and Andreas Nehring

Received: 8 August 2021 Accepted: 3 September 2021 Published: 8 September 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

of climate change), while being quite undisputed among scientists [11]. These topics are referred to as *societally denied science* [11] or *controversial science issues* (CSI) [12]. Attitudes toward CSI, i.e., their rejection or acceptance, highly rely on individual norms and values that do not necessarily result from scientific reasoning [13].

As science and technology develop rapidly, opportunities to encounter a variety of SSI, on which decisions must be made, and CSI, on which attitudes must be formed, become more frequent. Fostering the ability to make informed decisions on such complex issues and problems using evaluation and reasoning is not only a crucial aspect of general scientific literacy [14,15] but also a central goal of science education [16,17] and science communication [18,19]. Scientific reasoning [20,21] and informal reasoning (i.e., everyday reasoning on ill-structured problems) [22] on SSI and CSI entail the evaluation and justification of claims.

Several researchers have pointed out that argumentation is a core competence for reasoning and scientific inquiry [23], as well as central to science education in general [24]. There are few studies on argumentation in science communication [25], but it is a potentially beneficial field to bridge science communication and science education [18]. Argumentation in terms of SSI and CSI involves an ethical dimension, so socioscientific argumentation is a distinct process from scientific argumentation [26]. Multiple studies have demonstrated that reasoning on SSI [27,28] improves the complexity and quality of students' arguments concerning both scientific and socioscientific issues and can improve students' argumentation skills [29] and critical scientific literacy [30].

Different approaches to assessing argumentation have been used in science education [31]. *Toulmin's Argument Pattern* (TAP) [32] in particular has been applied in various ways [33]. However, TAP is predominantly used to assess the quality of students' arguments [34] and focuses on an argument's structure [35]. To date, few studies have examined the content of arguments [27,35] or justifications [36].

Furthermore, most research on informal reasoning and argumentation in the context of SSI and CSI focuses on either school [27–29,33,35] or university students [16,36,37], but similar analytical approaches to argumentations used by the public [18] could provide insights into controversial debates on scientific issues in everyday life. This research aims to describe different kinds of justifications used when people reason informally on selected CSI based on their attitude, using a fine-grained category system.

#### **2. Theoretical Background**

#### *2.1. Socioscientific Issues (SSI) and Controversial Science Issues (CSI)*

"Controversial Science Issues are scientific topics that, by their very nature, create discussions, debates, and questions because students are intrigued by these issues, question them or even have significant doubts about them" [12] (p. 26). Often, the description of controversy in the relationship between science and society is left implicit in science communication [38,39] as well as science education [36]. Borgerding and Dagistan [11] differentiate between three categories of CSI: *active science*, *societally denied* and *societally accepted science*, and SSI (Figure 1).

Controversies within active science are located within the scientific community itself (actual scientific frontier debates) [11]. Societally denied science refers to a negative attitude (i.e., rejection) toward scientific knowledge among the public (i.e., "x is not true"). This scientific knowledge was nevertheless generated within the scientific community and a scientific consensus on it exists [11].

**Figure 1.** Relationship between controversial science issues, active science, societally denied and accepted science, and socioscientific issues (adapted from [11]). **Figure 1.** Relationship between controversial science issues, active science, societally denied and accepted science, and socioscientific issues (adapted from [11]).

Both societally denied science and societally accepted science can serve as a foundation for SSI [11]. For instance, knowledge about stem cell research counts as societally accepted science [11], even if the application of this research can be addressed as controversial when teaching SSI. SSI are highly relevant to society and are often discussed controversially in the context of science education [5]. Examples of SSI include stem cell research, environmental issues and their possible solutions, and the creation of genetically modified organisms [40], and therefore, the applications of scientific knowledge in these areas [11]. Normative questions, like "Is the application of this technology just?" are addressed in typical SSI, reflecting the fact that SSI cannot be resolved by science and scientific inquiry practices alone [11]. SSI have an ethical dimension concerning the relationship between science, technology, and society [41] as well as a complex societal dimension [5]. Problems in the context of SSI are open-ended, ill-structured, subject to multiple perspec-Both societally denied science and societally accepted science can serve as a foundation for SSI [11]. For instance, knowledge about stem cell research counts as societally accepted science [11], even if the application of this research can be addressed as controversial when teaching SSI. SSI are highly relevant to society and are often discussed controversially in the context of science education [5]. Examples of SSI include stem cell research, environmental issues and their possible solutions, and the creation of genetically modified organisms [40], and therefore, the applications of scientific knowledge in these areas [11]. Normative questions, like "Is the application of this technology just?" are addressed in typical SSI, reflecting the fact that SSI cannot be resolved by science and scientific inquiry practices alone [11]. SSI have an ethical dimension concerning the relationship between science, technology, and society [41] as well as a complex societal dimension [5]. Problems in the context of SSI are open-ended, ill-structured, subject to multiple perspectives, and they lack clear solutions [37].

tives, and they lack clear solutions [37]. However, not all controversial scientific topics addressed in educational contexts fit this definition of SSI. Issues may be controversial and contested within the public sphere without being ill-structured and/or without lacking clear solutions. SSI are often described as inherently controversial [9,35] or as one kind of controversial issue [27,42]. Following the ideas that SSI are one kind of CSI and that a publicly contested issue is not necessarily denied by the public, we describe CSI as an umbrella term (see Figure 1), comprising different approaches of science communication and science education. When engaging in However, not all controversial scientific topics addressed in educational contexts fit this definition of SSI. Issues may be controversial and contested within the public sphere without being ill-structured and/or without lacking clear solutions. SSI are often described as inherently controversial [9,35] or as one kind of controversial issue [27,42]. Following the ideas that SSI are one kind of CSI and that a publicly contested issue is not necessarily denied by the public, we describe CSI as an umbrella term (see Figure 1), comprising different approaches of science communication and science education. When engaging in CSI, the question is not whether a certain issue is true or just but what the reasons for its controversy are.

CSI, the question is not whether a certain issue is true or just but what the reasons for its controversy are. Examples of CSI are evolution [43] and climate change [44], since parts of society doubt their theoretical scientific foundation, i.e., have a negative attitude toward them. Attitudes are conceptualized here as an affective assessment of an attitude object (e.g., evolution, climate change) [45]. Nevertheless, these topics do not lack a clear solution and Examples of CSI are evolution [43] and climate change [44], since parts of society doubt their theoretical scientific foundation, i.e., have a negative attitude toward them. Attitudes are conceptualized here as an affective assessment of an attitude object (e.g., evolution, climate change) [45]. Nevertheless, these topics do not lack a clear solution and are not ill-structured. For other CSI, such as vaccination [46] and GMOs [47], the controversy refers, at least in part, to the application of technology and touches the field of SSI.

are not ill-structured. For other CSI, such as vaccination [46] and GMOs [47], the controversy refers, at least in part, to the application of technology and touches the field of SSI. Moreover, different factors influence attitudes toward CSI topics, hence the distinc-Moreover, different factors influence attitudes toward CSI topics, hence the distinction between societally denied and societally accepted science. Most influencing factors that affect the rejection of scientific knowledge or applications are affective, such as emotions,

tion between societally denied and societally accepted science. Most influencing factors

ideology, or worldview, and are referred to as the roots of attitudes [13]. Attitudes toward vaccination depend on risk perception, barriers, trust, calculation, and responsibility for society [46], while factors like religious belief [43], trust in science, and knowledge about the *nature of science* (NOS) influence attitudes toward evolution [48]. Climate change attitudes are influenced by political identity [49] and an individualistic worldview [50], and attitudes toward GMOs are affected by views about natural purity [51] as well as emotions and intuitions [52]. These different factors illustrate the issue-dependency and high heterogeneity of predictors of the controversiality of a topic [53,54]. However, some factors seem to be general predictors of the acceptance and rejection of scientific knowledge, like conspiracy ideation [55] and knowledge about NOS [56].

#### *2.2. Informal and Scientific Reasoning*

Engagement in SSI often involves argumentation and decision-making processes that require reasoning processes, i.e., processes of building and evaluating arguments [57]. For a long time, research on reasoning focused on formal reasoning about well-defined problems [58] and followed a "deduction paradigm" [59]. However, it has been demonstrated that human reasoning is prone to biases, and everyday reasoning is in most cases informal reasoning [58]. Both formal (scientific) and informal reasoning are processes of generating and assessing arguments [60]. While the problems addressed in scientific reasoning are often well-defined and the respective premises are explicit, problems in informal reasoning are ill-structured and the premises are not always stated [61]. Informal reasoning tasks often involve generating and evaluating positions on complex issues that lack clear solutions [5]. However, the coordination of theory and evidence [4,60], as well as generating evidence-based justifications [60], is central in informal and scientific reasoning: "Foundational abilities that lie at the heart of both types of reasoning are the ability to recognize the possible falsehood of a theory, and the identification of evidence capable of disconfirm" [60] (p. 74). These abilities align with the epistemic dimension of scientific reasoning as described by Osborne [21].

As SSI typically involve contentious and open-ended problems, their negotiation and resolution can be characterized by informal reasoning [5,61], which is especially suitable for processes like decision-making about actions for which supporting and opposing arguments exist [57]. The ability to informally reason on SSI has been described as a crucial component of scientific literacy [5] and a central goal of science education [62].

In addition to components of scientific reasoning [20,21], reasoning on SSI requires the integration of societal and ethical aspects, also referred to as moral reasoning [63,64]. Sadler, Barab, and Scott [8] proposed the construct of *socioscientific reasoning* (SSR) to assess the reasoning practices associated with SSI. While research on SSR highlights the integration of ethical components that require moral reasoning [42], reasoning on CSI is not necessarily a matter of moral reasoning but a matter of personal attitudes and knowledge. This is because the questions concerning CSI are neither open-ended nor unsolvable dilemmas [11]. Therefore, frameworks developed to assess SSR competencies [65], decision-making on SSI [66], and SSI attitudes [67] cannot be applied to CSI in which a clear scientific consensus concerning scientific knowledge and/or its application has been reached. Assessing how people reason concerning their attitude toward a CSI asks for different approaches, e.g., the identification of informal reasoning types [37,61,68]. While some research results suggest that reasoning is consistent across different topics [65,69], other studies describe a topic-specificity [70,71].

#### *2.3. Argumentation Frameworks*

Argumentation is the communicative part of reasoning [22] and is addressed more and more by science curricula around the world [33]. Argumentation in science is an essential skill, not only for scientists and science students but also for citizens, to enable them to make informed decisions on (socio-)scientific issues in everyday life [33].

Argumentation in general can be described as an interplay of constructing claims or explanations and the corresponding evidence [32,72] to justify something [73]. A fine-grained conceptualization of argumentation has been an ongoing challenge for researchers, and a variety of frameworks exist [31]. Aside from differences among these frameworks focusing either on content [28], structure [74], or the epistemological quality [75] of arguments, all of these frameworks rely on Toulmin's Argument Pattern (TAP) [32]. The TAP builds a general structure of arguments (Figure 2) and a foundation to assess them [33]. Argumentation in general can be described as an interplay of constructing claims or explanations and the corresponding evidence [32,72] to justify something [73]. A finegrained conceptualization of argumentation has been an ongoing challenge for researchers, and a variety of frameworks exist [31]. Aside from differences among these frameworks focusing either on content [28], structure [74], or the epistemological quality [75] of arguments, all of these frameworks rely on Toulmin's Argument Pattern (TAP) [32]. The TAP builds a general structure of arguments (Figure 2) and a foundation to assess them [33].

**Figure 2.** Toulmin's Argumentation Pattern (adapted from [32,33]) and its application to a complex argument concerning the CSI of SARS-CoV-2. **Figure 2.** Toulmin's Argumentation Pattern (adapted from [32,33]) and its application to a complex argument concerning the CSI of SARS-CoV-2.

The *claim* of an argument is its conclusion. It is a statement of commitment [33] that every individual can agree or disagree with. The claim is based on several elements of the argument, with *data* representing the evidence for the claim being the central element. The data needs a *warrant* as a conclusive rule, turning the data into a relevant reason to support the claim. The warrant itself can furthermore be based on additional information called a *backing*. Because those three compartments form the justifying part of a persuasive argument [32] they are subsequently subsumed as the *justification* of a claim. The justification is opposed to the *rebuttal*, which contradicts it, and the *qualifier*, which describes the extent to which the justification allows valid conclusions. The *claim* of an argument is its conclusion. It is a statement of commitment [33] that every individual can agree or disagree with. The claim is based on several elements of the argument, with *data* representing the evidence for the claim being the central element. The data needs a *warrant* as a conclusive rule, turning the data into a relevant reason to support the claim. The warrant itself can furthermore be based on additional information called a *backing*. Because those three compartments form the justifying part of a persuasive argument [32] they are subsequently subsumed as the *justification* of a claim. The justification is opposed to the *rebuttal*, which contradicts it, and the *qualifier*, which describes the extent to which the justification allows valid conclusions.

As humans are easily capable of connecting statements in a logical way, the warrant is sometimes left implicit [76]. In the given example, the fact that people die from COVID-19 (i.e., data) can lead to the conclusion of SARS-CoV-2 posing a serious threat to human health without formulating the warrant (i.e., that the possible death forms a serious threat to human health). Equally, the data can be left implicit. Taking this into account, the articulation of a justification does not always include both data and warrant but sometimes appears as only one of the two components. As humans are easily capable of connecting statements in a logical way, the warrant is sometimes left implicit [76]. In the given example, the fact that people die from COVID-19 (i.e., data) can lead to the conclusion of SARS-CoV-2 posing a serious threat to human health without formulating the warrant (i.e., that the possible death forms a serious threat to human health). Equally, the data can be left implicit. Taking this into account, the articulation of a justification does not always include both data and warrant but sometimes appears as only one of the two components.

The TAP is often used as an analytical framework to evaluate argument quality [33]. When assessing arguments, an adapted version of TAP is often used. Qualifiers are often neglected to reduce the complexity [77–79]. The claim-evidence-reasoning approach is an established adaptation of the framework, in which, in addition to claim and evidence (i.e., data), warrant and backing are summarized as reasoning [79]. The TAP is often used as an analytical framework to evaluate argument quality [33]. When assessing arguments, an adapted version of TAP is often used. Qualifiers are often neglected to reduce the complexity [77–79]. The claim-evidence-reasoning approach is an established adaptation of the framework, in which, in addition to claim and evidence (i.e., data), warrant and backing are summarized as reasoning [79].

However, using TAP or its adapted forms as an analytical tool has also been criticized [33] due to the ambiguity of the arguments' elements [80] and the context-dependency of their interpretation [81]. In particular, differentiating between data and warrant, as well as warrant and backing, is difficult and depends on the context [77,82]. These challenges, as well as approaches that merge data and warrant [82], underpin the justification component (see Figure 2). However, using TAP or its adapted forms as an analytical tool has also been criticized [33] due to the ambiguity of the arguments' elements [80] and the context-dependency of their interpretation [81]. In particular, differentiating between data and warrant, as well as warrant and backing, is difficult and depends on the context [77,82]. These challenges, as well as approaches that merge data and warrant [82], underpin the justification component (see Figure 2).

Several other studies that assessed arguments did not rely on TAP but analyzed arguments dichotomously by focusing on one claim supported by a ground, i.e., a reason [28,36]. Often, *subjective* and *objective* justifications are distinguished [36,83]. A comparable distinction was provided by Jafari and Meisert [27], who distinguished between normative and fact-based reasoning. Objective justifications are sometimes further divided into *evidential* and *deferential* justifications [83], with deferential justifications appealing to an authority [83]. Justifications were found to be heterogeneous within a person's argumentation and to differ among different CSI [36].

Additionally, several studies have indicated the relation between knowledge about NOS and argumentation skills [84,85]. Studies on argumentations in the field of SSI predominantly focus on argument quality based on the TAP or adapted forms. However, when it comes to the argument's contents and the types of justifications within arguments, few studies are available [27,86].

#### *2.4. Research Questions*

Several researchers have pointed out that instruction and conceptual knowledge of argumentation can foster the use of more complex [28] and more fact-based [27] arguments in the science classroom. There are still societal debates on scientific topics that are not disputed in the scientific sphere and do not lack clear solutions, and these topics count as CSI. Scientific knowledge, or its application that is partly rejected by the public, points to negative attitudes toward a topic. As roots of such attitudes are known to be mostly affective [13], this leads to the question of how people justify these attitudes.

An assessment of justifications within arguments on CSI in the public sphere is a necessary first step to identify overall tendencies and context-dependencies in justifications as one element of informal reasoning. In the long run, a resulting framework may help equip students with the necessary skills to participate in these public debates.

Our study addresses the following research questions:


#### **3. Materials and Methods**

#### *3.1. Participants and Data Collection*

We conducted an online survey in German, distributed via social networks to reach the public in an informal learning context. Postings included a short introduction to the topic and targeted communities interested in CSI, e.g., through a comment on videos on genetically modified food (GMF; YouTube), anti-vaccine groups (Facebook), and science communicators (Twitter). This random sampling was justified by the aim to reach out to a heterogeneous sample and collect a wide range of different justifications on different CSI. Data were collected within a two-month period in summer 2020.

In total *N* = 398 volunteers took part in the survey, of which *N* = 265 completed the questionnaire up to the last page. Participation in the survey was voluntary and during free time, which might explain the high dropout rate. It was possible to skip questions or leave the survey at any point. All open answers were analyzed, regardless of if the data set was complete. For closed questions, listwise deletion was applied. The age of participants ranged from 16 to 85 with an average of 41 years. Participants from all 16 German provinces took part, and the education level ranged from high school students and people who left school without a degree to post-doctoral researchers.

#### *3.2. Instruments*

The research design was adapted from Lobato and Zimmerman [36]. Their survey included four CSI topics (evolution, climate change, GMF, vaccination) and involved confronting participants with a statement (i.e., claim) reflecting the scientific consensus on each topic (Table 1). We added a fifth statement on SARS-CoV-2 to the survey, as the pandemic has led to the most substantive large-scale, open, and public discussion of epidemiology and science in recent history [87]. The statements used by Lobato and Zimmerman [36] were modified whenever they seemed to express epistemological considerations that could also serve as justifications, like "Evolution is the best explanation" or "Medical research has demonstrated" (Table 1).

**Table 1.** Statements reflecting scientific consensus on five CSI topics (modified based on [36]). Statements reflect the claim in TAP [32].


Participants' attitudes toward the CSI topics (i.e., acceptance or rejection of the scientific consensus) were measured using a five-point scale to rate their agreement with the five claims. The participants were subsequently asked to justify (i.e., data/warrant/backing) their attitude on each claim in an open answer format and to think of possible reasons to change their position (i.e., rebuttals). In the following analysis, we focus on the justifications.

In addition, other potentially influencing variables were assessed: knowledge about NOS [88], religiousness [43,89], and conspiracy ideation [90]. The NOS measure focused on the tentativeness of scientific knowledge ("development" scale) with items like "New findings might change what scientists hold as true" [88]. The original seven items were reduced to six items. The scale measuring religiousness consisted of five items such as "I believe in God" [43,89]. The scale on conspiracy ideation [90] included items like "I think many important things happen in the world, which the public is never informed about" [90]. All of these scales measured agreement on a five-point rating scale.

#### *3.3. Data Analysis*

Results of all rating scales were merged to sum scores per scale. The rating items to assess attitudes toward claims concerning the five CSI topics were merged to one sum score for further analyses, representing attitudes toward scientific consensus.

Open answer format responses (i.e., arguments) were analyzed using qualitative content analysis [91] and the software MAXQDA Plus (VERBI Software, 2019, Berlin,

Germany). Components of the analysis are semantic units; every semantic unit was coded once. many). Components of the analysis are semantic units; every semantic unit was coded once.

many important things happen in the world, which the public is never informed about"

Results of all rating scales were merged to sum scores per scale. The rating items to assess attitudes toward claims concerning the five CSI topics were merged to one sum

Open answer format responses (i.e., arguments) were analyzed using qualitative content analysis [91] and the software MAXQDA Plus (VERBI Software, 2019, Berlin, Ger-

As a first step, based on TAP, we deductively derived an operationalization to identify the semantic units within respondents' arguments that can be categorized as justifications (Figure 3). This step was necessary since, even if the open answer format question concretely asked about justifications, some answers contained other argumentative elements or unrelated components. As a first step, based on TAP, we deductively derived an operationalization to identify the semantic units within respondents' arguments that can be categorized as justifications (Figure 3). This step was necessary since, even if the open answer format question concretely asked about justifications, some answers contained other argumentative elements or unrelated components.

*Educ. Sci.* **2021**, *11*, x FOR PEER REVIEW 8 of 20

*3.3. Data Analysis* 

[90]. All of these scales measured agreement on a five-point rating scale.

score for further analyses, representing attitudes toward scientific consensus.

**Figure 3.** Operationalizing Toulmin's Argumentation Pattern [32] to isolate the justification (i.e., analyzed unit) of the argument proposed in the open answer format as the first step of the analysis. **Figure 3.** Operationalizing Toulmin's Argumentation Pattern [32] to isolate the justification (i.e., analyzed unit) of the argument proposed in the open answer format as the first step of the analysis.

> If the semantic unit named reasons supporting the participant's position concerning the claim (e.g., "The risk of dying from the disease is higher than dying from the vaccine") it was coded as *warrant/data,* since those two argument components, as postulated by Toulmin [32], rarely appear explicitly as two distinct units. In this case, the warrant (i.e., "If the risk of dying from the disease is higher than dying from the vaccine, the vaccine is safe and effective") is left implicit, as is often the case in informal logic [76]. The conceptualization of Toulmin [32] also includes *qualifiers* influencing the magnitude of an argument (e.g., "If the vaccine is developed and tested responsibly") and *rebuttals* contradicting the conclusion (e.g., "Some people die from the side effects of vaccines"). For the fol-If the semantic unit named reasons supporting the participant's position concerning the claim (e.g., "The risk of dying from the disease is higher than dying from the vaccine") it was coded as *warrant/data,* since those two argument components, as postulated by Toulmin [32], rarely appear explicitly as two distinct units. In this case, the warrant (i.e., "If the risk of dying from the disease is higher than dying from the vaccine, the vaccine is safe and effective") is left implicit, as is often the case in informal logic [76]. The conceptualization of Toulmin [32] also includes *qualifiers* influencing the magnitude of an argument (e.g., "If the vaccine is developed and tested responsibly") and *rebuttals* contradicting the conclusion (e.g., "Some people die from the side effects of vaccines"). For the following analysis of the justifications, the rebuttals were merged with warrant/data as justifications (i.e., analytic unit; Figure 3), because the statement "Some people die from the side effects of vaccines" either justifies or rebuts a participant's position.

If the semantic unit was completely unrelated to the claim, it was coded as *unrelated*. If it was a restatement of the claim captured in the rating scale (e.g., "I think vaccines help") it was coded as *claim*. If the semantic unit was unrelated to the initial statement, e.g., the safety and effectiveness of vaccines, but still related to the issue (e.g., "No one should be forced to be vaccinated") the unit was coded as *problematization*. Semantic units that referred to the initial statement without using any argumentative component (e.g., "Why would I answer that?") were coded as *refusals*.

The first deductive coding step resulted in a majority of answers justifying the statement, as intended in the open question (Table 2). There was no evidence of structural differences between stated argument components for or against scientific consensus. The stated argument components did not depend on the attitude measured.



The semantic units identified as justifications underwent a second qualitative content analysis to build up the deductive-inductive category system and answer the research questions. Therefore, we started by gathering similar content in fine-grained subcategories and subsequently generalized the categories more and more [91] based on those presented by Lobato and Zimmerman [36]. In this way, it was possible to categorize the justifications based on content and build types of justifications on CSI topics. To improve the objectivity of our category system, a different researcher conducted a second coding on 30 complete data sets (11.3% of complete data sets) [92]. These double coded data sets are a representational sample to encompass the spectrum of the material. Cohen's kappa indicates a substantial intercoder agreement (κ = 0.68) [92]. Based on a discursive analysis of the coding results, codings were discursively changed when coding errors were identified. This led to increasement of Cohen´s kappa (κ = 0.84) and a refinement of coding descriptions.

The amount and proportion of coded semantic units per type of justification were calculated and compared across the five topics. To analyze relations between types of justifications and other variables, correlations were calculated.

#### **4. Results**

The claims reflecting the scientific consensus on the five CSI topics were generally accepted, representing a positive attitude toward these topics. The most accepted claim was evolution (95.3 % agreement), followed by climate change (87.6%), vaccinations (86.0%), SARS-CoV-2 (82.6%), and finally GMF (57.5%), the most contested claim (Table 3).

Figure 4 displays the deductively-inductively built fine-grained category system to distinguish different types of justifications on CSI. The categorization resulted in five types of justifications, with 25 subcategories. A justification that cannot be falsified or is dependent on individual beliefs belongs to the *subjective* type, and every other justification is *intersubjective*. Subjective justifications refer to normative statements that are grounded in values and beliefs (e.g., *ideology:* "God created all living beings"; *naturalistic fallacy:* "This is not safe, because it is not natural"; *argumentum ad hominem:* "Virologists are not trustworthy"). Intersubjective justifications were further distinguished into those referring to specific data to support the claim (*evidential*) or referring to a third entity as an authority (*deferential*). The mere mention of "evidence" did not count as an evidential justification but

was categorized as a reference to a body of knowledge and therefore as deferential. The determining differentiation between evidential and deferential justifications was their specificity; while deferential justifications refer to a rather general body or lack of knowledge about the topic, evidential justifications are quite focused on the single CSI.


**Table 3.** Frequency (proportion) of acceptance of scientific consensus concerning each CSI topic.

**Figure 4.** Category system of justification types concerning attitudes toward claims on CSI topics. Number of subcategories per category is in square brackets. **Figure 4.** Category system of justification types concerning attitudes toward claims on CSI topics. Number of subcategories per category is in square brackets.

All identified justification types were identified across all five topics. However, some justification types were more common depending on the particular CSI topic addressed. While subjective, empirical, and theoretical justifications tended to be rather topic-specific, deferential justifications appeared with a similar frequency across most of the topics (Table 4). Therefore, references to a body or lack of knowledge were used quite similarly across the different CSI. However, on the safety of GMF, the most contested statement, a comparably high number of justifications refer to a lack of knowledge. In contrast, the claims with the highest acceptance rates, i.e., anthropogenic climate change, evolution, and vaccination, were more frequently connected with justifications referring to third entities or vaguely defined bodies of knowledge such as "studies" or "evidence". Compared with theoretical justifications, empirical justifications were far more com-Deferential justifications were further divided into justifications referring to *a body of knowledge* (e.g., *science/research:* "That was proven by science"; *consensus:* "Almost all scientist agree on it"; *control mechanisms:* "There is a strict and transparent approval procedure for vaccinations") or a *lack of knowledge* (e.g., *no (sufficient) evidence:* "We don't know enough about it"; *no falsification:* "To date, there is no evidence against it"; *no personal knowledge:* "I don't know enough about this"). The evidential justifications were categorized as either *empirical* or *theoretical* justifications. While empirical justifications referred to verifiable real-world phenomena (e.g., *causality*: "As shown by the eradication of smallpox"; *comparison/analogy:* "SARS-CoV-2 is not more dangerous than the flu"; *definition/generalization:* "This is the case, since we have a global pandemic"), the theoretical justifications drew conclusions, weighed up, or referred to conclusiveness (e.g., *weighing up:* "The risk of dying from the sickness is higher than dying from the vaccine"; *conclusiveness*: "This is a conclusive explanation").

mon. However, this varied across the topics; while justifications concerning evolution relied almost equally on theoretical considerations and real-world observations, positions on SARS-CoV-2 were more frequently justified by empirical justifications. Subjective justifications were the least common justification type, with the topic of GMF showing the highest proportion of subjective justifications, while almost no respondents gave subjective justifications in the contexts of evolution and anthropogenic climate All identified justification types were identified across all five topics. However, some justification types were more common depending on the particular CSI topic addressed. While subjective, empirical, and theoretical justifications tended to be rather topic-specific, deferential justifications appeared with a similar frequency across most of the topics (Table 4). Therefore, references to a body or lack of knowledge were used quite similarly across the different CSI. However, on the safety of GMF, the most contested statement, a

change. The most frequent justification type overall was reference to a body of knowledge.

Subjective 0.4% 1.4% 7.3% 4.3% 3.6% 66

Body of knowledge 49.1% 55.2% 25.3% 45.9% 26.1% 644

Lack of knowledge 9.1% 5.9% 38.0% 6.5% 13.8% 230

Theoretical 17.1% 6.2% 3.0% 12.3% 1.5% 132

Evidential: 20.6% 31.4% 26.3% 31.1% 55.0% 526

**Change GMF Vaccination SARS-CoV-2** *N***Total**

**Table 4.** Proportions of justification types across the CSI topics.

**Justification Evolution Climate** 

change, evolution, and vaccination.

**Type of** 

Deferential:

Deferential:

Evidential:

comparably high number of justifications refer to a lack of knowledge. In contrast, the claims with the highest acceptance rates, i.e., anthropogenic climate change, evolution, and vaccination, were more frequently connected with justifications referring to third entities or vaguely defined bodies of knowledge such as "studies" or "evidence".


**Table 4.** Proportions of justification types across the CSI topics.

Compared with theoretical justifications, empirical justifications were far more common. However, this varied across the topics; while justifications concerning evolution relied almost equally on theoretical considerations and real-world observations, positions on SARS-CoV-2 were more frequently justified by empirical justifications.

Subjective justifications were the least common justification type, with the topic of GMF showing the highest proportion of subjective justifications, while almost no respondents gave subjective justifications in the contexts of evolution and anthropogenic climate change. The most frequent justification type overall was reference to a body of knowledge. This type was especially common when justifying attitudes on anthropogenic climate change, evolution, and vaccination.

In most cases, the acceptance of claims concerning the five different CSI topics did not correlate significantly with the identified type of justification (Table 5). However, the use of subjective justifications is negatively correlated to the acceptance of four of the CSI topics with a weak effect. The claim about GMF is the only one without a significant correlation to one of the justification types. Additionally, the acceptance of the effectiveness and safety of vaccines is significantly and weakly related to the use of deferential justifications referring to a body of knowledge.

**Table 5.** Correlation after Pearson between justification type and acceptance of scientific consensus on each topic. *N* = 398, \* *p* < 0.05, \*\* *p* < 0.01.


In general, the participants were not very religious (*M* = 1.62; *SD* = 1.05, score range: 1–5), were partly drawn to conspiracy theories (*M* = 2.41; *SD* = 0.88, score range: 1–5), and showed a high knowledge about NOS (*M* = 4.69; *SD* = 0.44, score range: 1–5).

A significant positive and strong correlation between the general acceptance of scientific consensus and knowledge about NOS (*N* = 252; *r* = 0.558; *p* < 0.01) was identified. Religiousness (*N* = 254; *r* = −0.469; *p* < 0.01) and conspiracy ideation (*N* = 258; *r*= −0.655; *p* < 0.01) correlated significantly negatively with the acceptance of scientific consensus with a medium (religiousness) to strong (conspiracy ideation) effect size.

Correlations of these variables with different types of justification were not significant in most cases (Table 6). Solely the use of subjective justifications (e.g., natural fallacy) correlated positively and weakly with the rejection of scientific consensus as well as negatively and weakly with conspiracy ideation. References to a body of knowledge correlated with the acceptance of scientific consensus with a weak effect. Furthermore, religiousness correlated weakly with the use of empirical justifications.

**Table 6.** Correlations between justification type and knowledge about NOS, religiousness, conspiracy ideation, and general acceptance of scientific consensus operationalized by the mean of acceptance of the claims on the five CSI topics. *N* = 398, \* *p* < 0.05, \*\* *p* < 0.01.


#### **5. Discussion**

The relatively high agreement with the claims on the different CSI indicates that most citizens who responded to the survey accept the respective scientific consensuses. However, while evolution as the explanation for the variety of life forms is accepted by more than 95% of the sample, only 57.5% agreed with the safety of GMF, the claim with the highest frequency of rejection and uncertainty. About 10% disagreed with the claims about the effectiveness and safety of vaccines and the health threat of SARS-CoV-2. Analysis of justifications resulted in five types of justifications for claims on CSI, each with several subtypes (RQ1). Justification types seem to be partly topic specific (RQ2) and in most cases are unrelated to whether the claim on a CSI was accepted or rejected (RQ3), as well as to variables like NOS, religiousness, and conspiracy ideation (RQ4).

#### *5.1. Justification Types in the Field of Controversial Science Issues (RQ1)*

To identify types of justifications in the field of CSI, we applied a deductive-inductive approach based on an existing justification coding scheme [36]. We identified subjective justifications that have been described before [36], sometimes referred to as normative justifications [27]. This type relies on individual spiritual, political, or ideological beliefs as well as on reasoning fallacies like *argumentum ad hominem*.

All justifications that could be identified as intersubjective formed a group that was further categorized. The distinction between references to a third entity (i.e., deferential) and references to the subject of discussion itself (i.e., evidential) was drawn from previous research [36] and applied to the data in this study. This common distinction can also be found in Shtulman [83].

However, taking a closer look at the deferential justifications, we distinguished references to a body of knowledge (e.g., "There is evidence for x") from references to a lack of knowledge (e.g., "There is no evidence"). Another step toward more fine-grained categories was the distinction within the evidential category between empirical and theoretical

justifications. Empirical justifications rely on real-world phenomena or precisely named and therefore provable data (e.g., correlation: "There is a positive correlation between greenhouse gas emissions and rising global temperature"), while theoretical justifications include a warrant to support the conclusion (e.g., cost risk calculation: "Even if climate change is not anthropogenic, we should assume it is. Better safe than sorry"). This categorization of the evidential justifications as either empirical or theoretical is therefore aligned with the distinction between data and warrant in TAP [32]. Furthermore, both types of evidential justifications share commonalities with components of scientific reasoning, e.g., the subskill of interpreting data [20] or abductive reasoning [93]. It would be worth investigating to what extent these types of evidential justifications align with the epistemic dimension of scientific reasoning as described by Osborne [21], referring to the questions "How do we know or how can we be certain?" [21] (p. 270).

Clearly, evidential justifications that refer to the CSI topic under consideration itself are highly topic-dependent. Due to the high diversity of SSI [70], a further generalization of this type of justification is challenging. One step that enabled the categorization into justification types was the focus on CSI as a special variant of SSI. Following Kolstø [94], who defined the field of risk-based SSI, and Borgerding and Dagistan [11] (see Figure 1), who differentiated between different fields as foundations for SSI, the theoretical clarification of the field of CSI as well as the resulting category system may help to further clarify the different fields within the broad topic of SSI. This is likely necessary for a finer analysis of justifications that could perhaps be field-specific.

### *5.2. Topic-Specific Justifications (RQ2)*

Despite the field-specific scope of the category system, indicated by the occurrence of all five justification types in all five CSI topics, the results show frequency differences among justifications concerning the five CSI. This finding supports earlier results with a similar methodological design [36], while results of studies investigating SSR instead suggest consistency of the SSR framework across different SSI contexts [8,69]. However, this contrast may be resolved by seeing the SSR framework as a field-specific tool that is applicable to different topics of SSI, comparable with the category system for the field of CSI presented here. Toulmin [32] has already emphasized the field-specificity of arguments.

Whereas subjective and evidential justifications appear to be more topic-specific, the most general justification types seem to be deferential justifications referring to a body or lack of knowledge, either personal or related to the scientific field. In fact, the vast majority of deferential justifications refer to the scientific field. However, as the participants were aware that they were part of a scientific survey, they may have tried to use appropriate and convincing arguments. Laypeople are often capable of using "public scientific arguments" [25].

#### *5.3. Relationship between Acceptance of CSI and the Use of Different Justifications (RQ3)*

Generally, correlations between the use of certain justifications and the acceptance of the scientific consensus on the different CSI were weak. Still, the use of subjective justifications correlated with a rejection of the scientific consensus on most CSI, except for the safety of genetically modified food. One possible explanation is that all kinds of fallacies (i.e., argumentum ad populum, argumentum ad hominem, naturalistic fallacy) are subjective justifications. This fallacious argumentation is known to be rather common when arguing against a scientific consensus [95]. Despite the only small number of subjective justifications in total, these correlations suggest that subjective justifications are more frequently formulated if people reject the scientific consensus on a CSI.

Deferential and evidential justifications seem to appear for both acceptance and rejection of the scientific consensus, indicated by the insignificant correlations between the use of these justifications and acceptance of the scientific consensus on the five CSI. The only exception is a significant and weak correlation between reference to a body of

knowledge and acceptance of the effectiveness and safety of vaccinations, indicating less frequent use of this argument when being skeptical about vaccinations.

#### *5.4. Relationship between NOS, Religiousness, and Conspiracy Ideation with the Use of Different Justifications (RQ4)*

Concerning the relationship between justification types and other variables, increased knowledge about NOS did not correlate with a certain type of justification, an observation made previously concerning NOS and the structural quality of arguments [47]. Nevertheless, NOS is known to be able to positively influence argumentation skills on SSI topics [85,96].

While Lobato and Zimmerman [36] noted that justification strategies appear highly heterogeneous within an individual's argumentation, our research demonstrated that even across the spectrum of science rejection and acceptance, all different kinds of justifications appear. This is consistent with previous findings that point out similarities in argumentation on supernatural beliefs and scientific knowledge [83]. However, significant correlations indicate that reference to a body of knowledge is more likely when accepting the scientific consensus, while subjective justifications are more frequent in argumentations against the scientific consensus.

Furthermore, subjective justifications are more frequent in people with high conspiracy ideation. Religiousness correlated weakly and positively with the use of empirical justifications, suggesting that religiousness is not necessarily an obstacle to reasoning on scientific topics [43].

#### **6. Conclusions and Outlook**

The task of fostering reasoning and argumentation competency goes beyond formal education in school and university [4]. In general, citizens are expected to employ evidencebased reasoning on issues grounded in science to make decisions in their personal lives and in public policy [97]. People often have difficulty evaluating evidence, which is problematic for informal reasoning on public policy and personal choices [4]. One crucial reason that these everyday reasoning tasks are difficult is the easy generation of causal explanation and their resistance [4,98].

To equip citizens with the ability to weigh up arguments and evaluate evidence, a first step is knowledge about the different types of justifications they provide for their attitudes concerning certain CSI. The category system reflecting justification types provides insight into the diversity of argumentation patterns and can inform teachers and pre-service teachers about potential attitudes and justifications on CSI that they might encounter in their lessons. Previous studies emphasized the importance of the inclusion of multidisciplinary perspectives when negotiating complex societal issues like CSI [7,35]. This approach can be informed by the category system, which was built upon a wide variety of different justifications from a heterogeneous online sample. It could furthermore be a helpful tool for fostering *science media literacy*, described by Höttecke and Allchin [99] as a crucial goal of science education in the age of social media [99].

Moreover, the presented category system lays the groundwork for further research in this area. On one hand, it will be the starting point for similar research in formal education. On the other hand, knowledge about justification types and how they differ across different contexts enables the ability to choose the best contexts to integrate into science education contexts.

Additionally, the results may inform science communication researchers and practitioners about the acceptance of the scientific consensus on different CSI topics and common justifications in these contexts. This is important, since even media reports often have problems handling scientific information [19].

In future research, the fine-grained assessment of general attitudes toward SSI brought forward by Klaver and Walma van der Molen [67] could be combined with the method of measuring justifications toward scientific consensus on specific CSI proposed in this article to shed more light on the different justification types. Furthermore, a research design integrating a task on SSR would be beneficial, e.g., by using the QuASSR [65]. In general, further investigation of the category system and its justification types should include steps of further validation [100] as well as argumentation in a broader discussion context, as has been suggested by several scholars [32,33,72]. The current study involved a random sample recruited within social networks to collect a wide variety of justifications for creating the category system. However, this sampling led to a high dropout rate and lacks representativeness of the quantified results. Future studies may apply the category system to samples within controlled environments. Another important next step is the theoretical and empirical investigation of the alignment of scientific reasoning and informal reasoning on CSI and SSI.

The novel term CSI could—following further theoretical and empirical clarification help bridge the gap between the mostly separated research areas of science education and science communication [18].

**Author Contributions:** Conceptualization, A.B.; methodology, A.B.; validation, A.B., L.M. and A.U.z.B.; formal analysis, L.M.; investigation, L.M.; resources, A.B. and A.U.z.B.; data curation, LM. and A.B.; writing—original draft preparation, L.M. and A.B.; writing—review and editing, A.B., A.U.z.B. and L.M.; visualization, A.B.; supervision, A.B. and A.U.z.B.; project administration, A.B. and A.U.z.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Informed Consent Statement:** The respondents agreed to data use for research.

**Data Availability Statement:** The datasets are not publicly available.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Valeria M. Cabello 1,2,\* , Patricia M. Moreira <sup>1</sup> and Paulina Griñó Morales 2,3**


**Abstract:** Constructing explanations of scientific phenomena is a high-leverage practice that promotes student understanding. In the context of this study, we acknowledge that children are used to receiving explanations from teachers. However, they are rarely encouraged to construct explanations about the causes and consequences of phenomena. We modified a strategy to elicit and analyze primary students' reasoning based on scientific theory as a methodological advance in learning and cognition. The participants were fourth-graders of middle socioeconomic status in Chile's geographical zone with high seismic risk. They drew explanations about the causes and consequences of earthquakes during a learning unit of eighteen hours oriented toward explanation-construction based on the Tectonic Plates Theory. A constant comparative method was applied to analyze drawings and characterize students' reasoning used in pictorial representations, following the first coding step of the qualitative Grounded Theory approach. The results show the students expressed progressive levels of reasoning. However, several participants expressed explanations based on the phenomena causes even at an early stage of formal learning. More sophisticated reasoning regarding the scientific theory underpinning earthquakes was found at the end of the learning unit. We discuss approaching elementary students' scientific reasoning in explanations based on theory, connected with context-based science education.

**Keywords:** explanations; scientific reasoning; drawings; science education; earthquakes

## **1. Introduction**

Instructional practices that are central to learning are called high-leverage practices [1]. Constructing explanations based on evidence derived from inquiry processes [2] or underpinned by scientific theories or principles is relevant for mobilizing students' understanding of natural phenomena in science classrooms [3].

Constructing better explanations continuously provides an organizational and educational framework for designing science teaching and learning experiences [4]. Elementary school students' explanation construction has been researched primarily in developed countries, i.e., [5]. Nonetheless, in developing countries, this field of research is in its early years [6]. Moreover, most of the studies in elementary classrooms are based on students' written explanations [6]. For instance, Forbes et al. [5] found that German classrooms supported students' use of evidence to ground claims. At the same time, the teachers gave more robust opportunities to evaluate evidence-based explanations through comparison in the US. Hence, students learned to look for bias in their reasoning by analyzing other students' explanations. In primary school, exploring and fostering students' explanation construction at the same time is difficult because the students are at the entry points to learn the theories, concepts, or principles. They also start developing writing skills and knowing to use the diverse genres in science education [6]. Thus, misinterpreting students' knowledge because of them having diminished writing skills is likely to happen.

**Citation:** Cabello, V.M.; Moreira, P.M.; Griñó Morales, P. Elementary Students' Reasoning in Drawn Explanations Based on a Scientific Theory. *Educ. Sci.* **2021**, *11*, 581. https://doi.org/10.3390/ educsci11100581

Academic Editors: Moritz Krell, Andreas Vorholzer and Andreas Nehring

Received: 8 August 2021 Accepted: 21 September 2021 Published: 26 September 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Despite the different emphases of the international studies, there is agreement that explanation-construction is a challenging task for students and teachers which requires explicit support from linguistic and conceptual areas [6,7] or distributed scaffolding to help students gradually [8,9]. Linking phenomena with their underlying causes appears to be among students' difficulties in constructing explanations. This process requires scaffolding and reframing the thinking mechanisms to include theories or concepts already existing in the individual's system of theories [9–12]. Indeed, there is a need for research on evaluative approaches to scaffold students' construction of scientific explanations [11].

Scientific explanations constitute a specialized genre of the discipline in the classroom, different from the report, arguments, or other text genres that children might be more familiar with [13,14]. Constructing explanations also involves the development of causal reasoning [15,16], disciplinary-specificities, i.e., [17], and the transformation of the individuals' intuitive theories [18]. This transformation is influenced by formal knowledge [19], which usually occurs in a social dimension of learning in the classroom activity. Additionally, scientific reasoning skills and other cognitive, metacognitive and motivational—social—skills are related to one another [11,20]. Managing all these dimensions is relevant but also challenging for teachers and researchers when they identify the development of the explaining practice and engage primary students in making sense of phenomena [20–24].

The current study focuses on analyzing pictorial representations of a specific phenomenon, earthquakes, in elementary school students to understand better the process of eliciting their causal reasoning through drawn explanations during a learning sequence. The objective was to characterize students' expressed reasoning through drawn explanations. Using drawings for this purpose advances an evaluative approach to younger learners' thinking, who are just learning to write and talk in science. Additionally, analyzing drawings complements the classic methodological trends of verbal and written modes of making meaning. This knowledge is needed to analyze students' reasoning in phenomena underlined by a scientific theory and identify alternative formats to benefit the growing number of students learning science through a foreign language or those with verbal/oral expression difficulties [24].

#### *1.1. Explanation-Construction as a Meaning-Making Process*

Creating or sharing meaning in science education involves multimodal languages, experiences, and interactions in the classroom [21]. The students' construction of explanations as a source of expressing their ideas is crucial, as it provides a window to understanding and sensemaking [22]. A teaching approach responsive to meaning-making processes will anticipate students' ideas about phenomena before instruction and then elicit and respond to these ideas during the lesson [23]. The materials and resources are other crucial elements for meaning-making processes [24].

From a sociocultural perspective, students' explanation construction is a strategy for knowledge integration. It is an iterative and collaborative process in which they connect what is already known—by prior instruction or intuitive theories—with their experiences and conceptual elements to give scientific support for certain phenomena. From this perspective, the explanations constructed are learning artifacts rather than products or learning samples [25]. Explanations in the form of pictorial representations are considered in this study to be vehicles for thought, or reasoning artifacts [26] that trigger the creation of meaning [21] and, consequently, turn into steps in the development of precursor models. These are "cognitive schemata compatible with scientifically appropriate knowledge since they are constructed on the basis of certain elements pertinent to scientific models, which have a limited range of application, and which prepare children's thinking for the construction of scientifically appropriate models" [27] (p. 2259).

In cognitive terms, explanation-construction requires a process of reasoning about phenomena [17] that is rarely easy to access as an external observer since it might require the recreation of the "image of the world" of the other, which contains not only concepts, but the images created through visual thinking [28]. Indeed, even when teachers know their students' initial ideas, it is hard to build on those ideas while teaching to probe their students' reasoning [29].

Even though interpreting and building new ideas based on students' reasoning in the classroom is challenging for teachers, encouraging the students to construct explanations provides an optimal scenario to engage in understanding natural phenomena, such as those related to socio-scientific issues [30,31]. Moreover, these scenarios help them reconstruct their knowledge and reasoning about phenomena relevant to their lives [32,33]. The reasoning process elicited in the classroom is afforded by an interaction between two information processing systems: the individual's intuitive and deliberative thinking [19]. Categories, as hypothetical entities in science education, fall under the umbrella term of "concept". These entities are products of reasoning with theoretical inputs provided by formal education [34]. We understand explanations as a vehicle for triggering learning and expressing scientific reasoning that emerges when putting the ideas into a material form of communication (see the next section). Therefore, we interpret students' drawings from the lenses of sociomateriality, both as processes that elicit reasoning and as outcomes of expressed scientific reasoning about a phenomenon that appears to be of high risk.

#### *1.2. Explanations and Students' Scientific Reasoning*

Children at school learn about the underlying principles of phenomena and causal relationships, usually but not exclusively in science education. These learning processes are crucial to developing scientific thinking, which is applying the methods or principles of scientific inquiry to reasoning or problem-solving situations [35]. We understand scientific reasoning from a multiple component skills perspective [36], including hypothesizing, experimenting, and evaluating evidence (inferencing, evaluating data, and drawing valid conclusions) [37]. Generating valid conclusions in inquiry processes usually requires explanations. Explanations are particularly characteristic of everyday causal understanding appearing during early childhood [16].

This article studies a specific component, causal scientific reasoning expressed or demonstrated in children's explanations [16] if we take them as a process of intentional knowledge-seeking [36]. Causal scientific reasoning emerges when they need to explain why a specific phenomenon occurs. Constructing explanations requires diverse causal connections [38], which means identifying particular circumstances that can trigger consequences to understand why observed changes or phenomena have a place under certain conditions. Explanations in science education involve scientific knowledge, and they can be based on theory, evidence, and mixed with daily life experiences. Children's scientific reasoning reconciles different kinds of causal explanations about phenomena, such as scientific, natural, and supernatural [17].

Explanations in science education frequently involve abstract knowledge or concepts (i.e., explaining phenomena at an atomic or molecular level mediated by energy transfers). Into a framework for modeling competence, explanations in science classrooms trigger children's abductive reasoning, which is the theory-based attempt of explaining a phenomenon by a cause [38]. Abduction means generating a cause as the best explanation for a phenomenon based on theoretical knowledge [39].

Considering scientific reasoning components, children's use of information to make causal inferences is a complex cognitive task [35,40]. However, this does not imply that young learners cannot express causal reasoning about their natural environment [41]. Wang et al. [42] observed how children between 2 and 5 years old faced causal tasks related to the weight of objects and concluded that, even before primary school, children use causal reasoning in natural environments, although some age-dependent variations were found. Mayer and collaborators [20] measured four scientific reasoning dimensions in everyday situations, one of those was understanding theories. They worked with 155 fourth-grade students in a paper and pencil instrument test. The results showed that children developed their performance in the measured dimensions.

In terms of searching for explanations to make sense of a phenomenon, scientific reasoning is related to the construction of models. A model used for teaching and learning concepts serves as a medium for communication, describing, and explaining [39]. Perkins and Grotzer [40] proposed a selection of causal models based on the level of reasoning sophistication: (a) mechanism, where students can use their experience to make generalizations not always aligned with mechanistic reasoning, moving to more complex and accurate explanations; (b) interaction pattern, a dimension where students use different paths to connect causes and effects; (c) probability, referred to as what could happen; (d) agency, for example when students identify the presence of an agent involved in direct action. Within each of these dimensions, the authors note sublevels of complexity. Based on Perkins and Grotzer's framework and other research studies of causal reasoning in science education, Moreira et al. [9] found that secondary students use complex causal reasonings to develop explanations in a specific chemistry topic. However, their results showed that using mechanistic reasoning does not always guarantee an alignment with scientific theory. Zangori et al. [31] built a rubric based on Perkins and Grotzer's framework [40] and other studies related to reasoning about ecosystems to analyze the causal associations used by third-grade students when they learn about ecosystems. They found the students who had the opportunity to reason using models enhanced their causal reasoning, and intermediate steps towards the use of causal reasoning were identified.

#### *1.3. Scaffolding Explanations in Science Learning*

Other studies have developed instructional models or learning progressions to scaffold, assess, and analyze students' explanations at the school level, e.g., [43,44]. These studies have common characteristics; they describe the explanation components and using evidence in their performances. McNeill et al. [43] constructed their instructional model considering Toulmin's framework and standards for science education, describing three explanation components: claim, evidence, and reasoning in the following components:

Claim, an assertion or conclusion that answers the original question; evidence, scientific data that support the claim; the data need to be appropriate and sufficient to support the claim; and reasoning, a justification that links the claim and evidence and shows why the data count as evidence to support the claim by using the appropriate and sufficient scientific principles.

However, a few studies relate explanations and scientific reasoning in evaluative purposes, for instance, highlighting the reasoning expressed by students in their productions. A five-stage comprehensive learning progression of written scientific explanations for the school level was designed by Yao and Guo [44]. At the more basic stages, the students first relate, indirectly, facts and theory through models. When their scientific reasoning evolves, they progressively approach scientifically accepted models. The elements of reasoning appear as a simple causality, moving forward to more complex forms such as probabilistic or correlational reasoning to link the explanations logically [44].

The distinction between school explanations based on evidence versus those based on theory is an ongoing academic discussion. However, we know that the scaffolding process that children need to construct explanations based on their observations, inquiry processes, and evidence is different from the practical support for students to create explanations underpinned by theories, principles, or models that are more abstract entities [43,45]. Among the first group, the studies show that systematically helping students distinguish between the description of the facts, observations, and the emergence of an argument based on evidence is worthy of learning, e.g., [43]. The second group of students' explanations supported by theories—counts with empirical support of how the use of epistemic tools, such as the Premise–Reasoning–Outcome instructional strategy (P.R.O.) [45] facilitates not only writing of better explanations but enhancing students' cognition and metacognition processes [46]. Thus, in the context of learning to explain phenomena based on theory, we found the research need of a domain-specific instrument to characterize students' reasoning and apply it to explanations.

Previous studies of explanations as a product and process of learning have analyzed verbal or written answers separately, i.e., [9,47,48]. However, this type of analysis has insufficiently captured the complexity and advancement of children's reasoning in learning new scientific concepts [47,49].

Consequently, we focus on generated pictorial representations in drawings, a complementary format vital for children's expression and communication that has been less researched in this field [49]. In addition, Park et al. [50] argued that this type of representation contains implicit information that offers an opportunity to analyze students' ideas and concepts. Indeed, analyzing non-linguistic forms of representation is a more inclusive method to approach students with difficulties with verbal/oral expression [24].

The focus of our work is highlighting and approaching children's reasoning about natural phenomena underpinned by theory from a cognitive perspective. We centered the application of this purpose on student-generated drawings as an alternative form of constructing and communicating explanations to make sense of the causes of a natural phenomenon that might affect their lives, specifically earthquakes. We chose the earthquake phenomena because, in Chile, the country in which this study was conducted, earthquakes are a relatively frequent event that children are familiar with, as the country is in a seismic area. Thus, for the participants living in a geological fault zone, this phenomenon might be more quotidian/frequent or, at least not as unfamiliar as other natural phenomena. Nonetheless, the fourth grade is the first formal opportunity in which students start learning the underpinning theory of this phenomenon, known as Tectonic Plates Theory (henceforth, TPT). Moreover, the transmutation of the daily life self-explanations of phenomena towards scientific explanations based on theory begins at the stage this research took place.

Briefly explained, TPT states that layers and plates form the Earth's internal structure in the static model. Plates move in different directions, giving place to continents as we currently know them. The inner movements of the plates occur mainly in three forms, convergent where plates move towards each other, divergent where plates move away from each other. Lastly, in transform movement, each plate moves sideways compared to the other. As a result of such movements, energy builds up, released through earthquakes, tsunamis, and related events. Therefore, TPT describes movements of plates, explaining the origin and mechanism of earthquakes [51].

We started from the assumption that supporting students in constructing explanations is a high-leverage practice in education [3], implying the development of reasoning processes and more authentic scientific practices in this study regarding TPT.

Our research question was: What characterizes students' expressed reasoning in drawn explanations in the context of learning about earthquakes? The purpose of this article is to shed light on primary students' scientific causal reasoning during a learning sequence at the school, in the context of current challenges in science, as well as to present a novel methodological coding rubric to approach this process. Science education needs to promote students' thinking processes through authentic scientific practices, such as constructing explanations. Thus, this work will contribute to research on primary students' causal reasoning and science education from a cognitive perspective.

#### **2. Materials and Methods**

The present study was exploratory with a descriptive and relational scope based on educational practices to inform educational processes. The data set was collected in Spanish and then translated into English by the article's first author for dissemination purposes. The information from the participants was gathered during the science learning sequence about the "Internal Dynamics of the Earth" in 2019. Two stages during the learning sequence were crucial for collecting the data that compose this study, part of a larger project in science education research. These stages are denominated as stage one and stage two, henceforth S1 and S2. S1 represented when the learning unit was started by the teacher, and S2 when the unit finished. It is important to note that this study did not intend to estimate the effectiveness of the teaching unit or determine how the learning opportunities

provided affected students' scientific reasoning skills because the study design did not include an intervention or comparison groups to make those inferences.

Characterizing students' drawings provides opportunities to analyze how instruction and the curriculum need to challenge students' ideas. It is educationally relevant considering that students' and scientific ideas coexist and interplay in their experience of making sense of the world [52]. The instruction helps with a reconstruction of these ideas in the sense of an explanatory coexistence [52].

The learning sequence in our study consisted of approximately 18 h of pedagogical work distributed throughout four weeks. The lessons comprised drawing activities, a group puzzle about Tectonic Plates and watching videos about the consequences of earthquakes, tsunami, and volcano eruptions. The teacher delivered some lectures about Earth Structure and Tectonic Plates' interaction. The students completed learning workbooks about the more dangerous hazards in Chile and socialized a school security plan.

During the learning unit, the learning outcomes were formalized by constructing hand-drawn explanations about the phenomenon of earthquakes. However, the teacher also used other sources to facilitate learning advances regarding tsunamis and volcanic eruptions. The prompt for triggering student drawings used in this study was "Why does the ground move (in a seismic context)? Please draw your explanation in this blank sheet". The instruments and steps of this study were approved by the Pontificia Universidad Católica de Chile's Ethics Committee code number 180514006.

#### *2.1. Participants and Paradigm*

The participants were 22 fourth-grade students from families of middle socioeconomic status. The school was selected through purposive sampling and was in an area of Chile identified as being at risk for disaster if an earthquake occurs, near the San Ramon geological fault line in Santiago, the Chilean capital. The partnership with the teacher for the educational purposes of this research included the collaborative design of a learning sequence to help students reason about the causes and consequences of Earth phenomena and, therefore, to construct scientific explanations through drawings. This decision was founded on the participatory research paradigm [53], in which the communities of research are part of the analytic process and the decision-makers in the study.

Although the whole class that composed the group participated in the learning activities, only 22 of the students had parental authorization and their consent to use the drawings for research purposes. Moreover, one student did not attend school the day the teacher allocated time for drawing in S2, and he did not want to do it later. Thus, the final data set consisted of 22 illustrations in S1 and 21 in S2, and some results are presented as percentages.

#### *2.2. Data Analysis and Processing*

Our data processing was carried out in three different steps. First, we developed an instrument to categorize the scientific reasoning expressed through drawn explanations following the study by Park et al. [50] about pictorial representations. Then, we used the constructed instrument to analyze a group of students' drawn explanations of earthquakes based on Tectonic Plates Theory (TPT). In the following paragraphs, we describe these two steps.

1. First, we developed an instrument to categorize the scientific reasoning expressed through drawn explanations following the study by Park et al. [50] about pictorial representations when qualitatively learning physics. Their work established three main levels for students' expression: sensory that includes what students sense; unseen substance level, which provides for concrete substances that cannot be seen; and lastly, unseen nonsubstance that contains those representations about non-concrete and unseen aspects. This prior work was developed with talented students, representing a novel contribution to the field with a limited scope of applicability.

A panel of three experts, including teachers and cognitive psychologists, checked that this first version of the instrument was conceptually adequate, and the levels proposed would be observable in regular primary students learning samples.

2. To expand the applicability and address explanations of regular primary students, we developed a first pilot qualitative analysis of a set of learning samples composed of drawings using the constant comparative method as the primary coding process of Grounded Theory [54]. We created groups of similar drawings and contrasted their main features, discussing the expressed reasoning that could be identified. Then, we went through three flows of activity of the constant comparative method to adjust the instrument to the data: data reduction, data display, and conclusion verification. We also followed the indications by Tang et al. [55] for interpreting specific aspects of children's drawing, such as types of lines for representing movement. Once we went through three rounds of discussion between the authors of this study, clarifications on the instrument were added. We modified the first version of the rubric by adapting the sensory level, the unseen substance level, and the unseen non-substance level of Park et al.'s framework [50], with specific emphasis on explanations of earthquakes based on TPT and an interpretation of the younger student's context-related scientific reasoning.

3. We conducted a qualitative analysis of students' explanations by three independent researchers—also authors of this article—all trained to code the drawings in a blind review process using the instrument developed in the previous steps. The final version of the rubric, which served as a coding framework, is presented in Table 1. The coding process was performed by each researcher independently; a total of 30% of the students' drawings were coded and compared among the three researchers in two rounds. The first round comprised 15% of the data, and the inter-rater reliability was 62%. After discussing the cross-cutting drawings, examples were selected to represent each level (see details in Section 3.3). The disagreements were discussed until a consensus was reached between the three researchers. The second round included a second set of drawings that comprised another 15% of the data set; the inter-coder agreement was 91%, which was considered a high measure of transparency for instrument implementation [56]. The remaining data were coded by one of the researchers considering the high level of prior agreement. The drawings were coded according to the three rows of the rubric. The first identified the main characteristics of the explanations represented in the students' drawings, looking for causes or consequences of the characterized phenomena. The second one centered the attention on the specific elements or details found in the representations. The third one interpreted the type of reasoning the student expressed in each drawing.



**Table 1.** *Cont.*


#### **3. Results**

This section describes first the coding framework and the rubric developed to characterize the students' expressed reasoning through drawn explanations. Secondly, we present the application results for fourth graders' drawings based on the main elements that constituted the participants' explanations based on theory. After this, we show the main trends of this group of participants' reasoning levels coded at the beginning and the end of a learning unit in context-based science learning related to earthquakes to illustrate a practical application of this novel approach. These results are presented as an example of the possible analysis of drawn explanations using the developed instrument but do not limit the application to one phenomenon only. Finally, we illustrate the composition of each reasoning level with some drawing examples, highlighting their inferior and superior anchor to orient teachers and researchers on the transitions from one reasoning level to the next one in the case of TPT.

#### *3.1. Instrument for Characterizing Scientific Reasoning in Drawings*

The instrument developed in our study takes the form of a comprehensive rubric which works as a coding system to facilitate the assignment of levels, and the characterization of primary students' expressed reasoning through drawn explanations. The rubric allows a description of both the characteristics of the domain-specific drawings and the reasoning level that might be externally interpreted.

Precisely, the rubric developed in this research (Table 1) consists of a three-level grid oriented to progressively identify levels of scientific reasoning in primary students, which are presented as columns. However, the first column represents a level 0 for drawings under the category of missing. As the instrument was applied to learning about earthquakes, its specification for Earth Science phenomena and TPT theory is included. We decided to base our work on distinguishing between perceptual planes expressed as input for interpreting reasoning and the connection between the explained phenomenon and its underpinning theory. This decision sought broader use of this approach to characterize early stages of students' scientific learning based on theories for modeling and explaining phenomena. In the topic of this study application, this stage corresponds to the fourth grade.

Additionally, the instrument added a minimum level used to code the learning samples that could not be categorized or did not answer the cognitive demand of the task, which is quite frequent in young children or during initial learning processes. We expected that students' drawn explanations move throughout the starting levels, from concrete or straightforward stages—based on their previous experiences, highlighting a sensorial focus—to more abstract ideas considering causal links, likewise expressing more complex reasoning. Furthermore, the rubric would make visible the sophistication of the students' expressed reasoning and understanding of the underpinning theory. Thus, the levels proposed in our instrument could also be used as an emergent learning progression.

The three rows of the rubric present elements as follows.

1. The first one describes the main characteristics of the explanations represented in the students' drawings, emphasizing the differentiation between their expressed sensory plane and the connection with the theory.

2. The second row presents the specific elements or details found in fourth-graders learning about a particular phenomenon, in this case, earthquakes, as an application of the first row to domain-specific learning samples.

The two first parts of the instruments may be adapted for working with other theories or phenomena.

3. The third row describes the interpreted scientific reasoning in connection with the sensory planes, the causality, and the usage of theory as a more abstract step in the students' cognitive processes when learning science. This part of the instrument is not associated with a singular theory; thus, it does not need adaptation to apply other topics.

The interpretation of reasoning is suitable to be used by educators or researchers in other learning topics or areas beyond Earth Science when students construct explanations based on scientific theories. It constitutes the first contribution of our work related to science learning research transcending the specific theory and expanding the cognitive process of causal reasoning rather than focusing on the learning accuracy of scientific concepts.

#### *3.2. Trends in the Participants' Reasoning Levels*

Considering the categorization results of the participants' explanations using the instrument described earlier (Figure 1), we observe that in the early stage in the formal process of learning—called Stage 1—(S1) before the learning unit began at the school, 28% of the students' explanations did not achieve the minimum level for categorization. Consequently, level 0 was assigned, as shown in Figure 1. In comparison, 24% of the drawings were categorized at level 1 for reasoning and 48% at level 2. This result means that most fourth-grade students could express reasoning about earthquakes with attempts to go beyond their immediate perception plane, representing elements that might constitute a causal explanation later, even with no formal instruction. However, none of the drawings reached level 3, causal reasoning based on aspects of TPT. Thus, we observed that some of them might have had an intermediate level of reasoning even with no formal instruction in this group of students. Nonetheless, establishing connections between the phenomenon and the theory in the form of a causal explanation in the drawings was difficult for the students. *Educ. Sci.* **2021**, *11*, x FOR PEER REVIEW 10 of 20 group of students could express more sophisticated reasoning and a causal link in their drawn explanations. Nonetheless, 45% of children did not show cognitive operations with unobservable entities or logically connect the causes and consequences of earthquakes, even after the learning unit was finished.

**Figure 1.** Participants' reasoning levels at the beginning and end of the unit. **Figure 1.** Participants' reasoning levels at the beginning and end of the unit.

or represent here your explanation about Why does the ground moves?"

tions categorized in this level in the current study are presented in Figure 2.

*3.3. Characterization and Examples of Reasoning Levels Interpreted from Drawings*  This subsection presents descriptions, main features, and examples for each level identified, representing the finest-grain analysis of student drawings. It is worth remem-In Stage 2—(S2), after the learning unit about the internal dynamics of the Earth was implemented, we saw a reduction in the percentage of drawings at level 0, with only 9% of the students' samples categorized as such. There was a proportional increase in level

bering that, in the context of learning about Earth Sciences, the task demanded was "draw

phenomenon from the pictorial representation. For instance, some students wrote "I don't know," drew a non-related phenomenon from an external observer's view, left the paper blank, or presented elements that were incomprehensible for the researchers in the light of the question demanded by the task. Thus, we could not interpret the students' expressed reasoning from these types of drawings. The authors of this work considered this level as missing data. This means that interpretable reasoning could not be obtained from an external viewer solely from drawings regarding the question given. However, other researchers might combine these types of illustrations with oral or written explanations; thus, the character of missing data would change. Some examples of pictorial representa-

(**A**) (**B**)

1 illustrations, with 36% classified as level 1 instead of the 24% obtained at S1. It is interesting to note that student representations categorized as level 2 decreased from 48% to 23% compared to S1; however, this fact is attributed to an increase in the drawings categorized in level 3, comprising 32% of the total. Thus, we conclude that, after participating in a formal learning process about earthquakes, it is likely that most of the participants in this group of students could express more sophisticated reasoning and a causal link in their drawn explanations. Nonetheless, 45% of children did not show cognitive operations with unobservable entities or logically connect the causes and consequences of earthquakes, even after the learning unit was finished. **Figure 1.** Participants' reasoning levels at the beginning and end of the unit.

group of students could express more sophisticated reasoning and a causal link in their drawn explanations. Nonetheless, 45% of children did not show cognitive operations with unobservable entities or logically connect the causes and consequences of earthquakes,

#### *3.3. Characterization and Examples of Reasoning Levels Interpreted from Drawings 3.3. Characterization and Examples of Reasoning Levels Interpreted from Drawings*

*Educ. Sci.* **2021**, *11*, x FOR PEER REVIEW 10 of 20

even after the learning unit was finished.

This subsection presents descriptions, main features, and examples for each level identified, representing the finest-grain analysis of student drawings. It is worth remembering that, in the context of learning about Earth Sciences, the task demanded was "draw or represent here your explanation about Why does the ground moves?" This subsection presents descriptions, main features, and examples for each level identified, representing the finest-grain analysis of student drawings. It is worth remembering that, in the context of learning about Earth Sciences, the task demanded was "draw or represent here your explanation about Why does the ground moves?"

Level 0: It is impossible to interpret an explanation connected with the earthquake phenomenon from the pictorial representation. For instance, some students wrote "I don't know", drew a non-related phenomenon from an external observer's view, left the paper blank, or presented elements that were incomprehensible for the researchers in the light of the question demanded by the task. Thus, we could not interpret the students' expressed reasoning from these types of drawings. The authors of this work considered this level as missing data. This means that interpretable reasoning could not be obtained from an external viewer solely from drawings regarding the question given. However, other researchers might combine these types of illustrations with oral or written explanations; thus, the character of missing data would change. Some examples of pictorial representations categorized in this level in the current study are presented in Figure 2. Level 0: It is impossible to interpret an explanation connected with the earthquake phenomenon from the pictorial representation. For instance, some students wrote "I don't know," drew a non-related phenomenon from an external observer's view, left the paper blank, or presented elements that were incomprehensible for the researchers in the light of the question demanded by the task. Thus, we could not interpret the students' expressed reasoning from these types of drawings. The authors of this work considered this level as missing data. This means that interpretable reasoning could not be obtained from an external viewer solely from drawings regarding the question given. However, other researchers might combine these types of illustrations with oral or written explanations; thus, the character of missing data would change. Some examples of pictorial representations categorized in this level in the current study are presented in Figure 2.

**Figure 2.** Examples of level 0. Drawing (**A**) shows a volcano, (**B**) represents the Earth planet and where Chili is.

Level 1: The student drawing represents elements within their sensory plane, generally as effects or consequences of the earthquake phenomenon, recognizable as movements of the ground's surface or results of the movement. The information derived from the representation was insufficient for the researchers to interpret an explanation beyond the child's perceptible plane, for instance, based on non-visible entities. These drawings frequently have a baseline to delimit the ground line (in a continuum, backstitch, or oblique) or function as object support. Some graphics also wrote words related to "movement" or "seism", etc., while others designed zigzag or wavy lines to represent the consequences of movement on the objects, as Figure 3 shows. Thus, we interpreted these drawings as a sensory level of reasoning because the cognitive operation is based on entities or elements within the students' perception of their senses.

where Chili is.

elements within the students' perception of their senses.

**Figure 2.** Examples of level 0. Drawing (**A**) shows a volcano, (**B**) represents the Earth planet and

Level 1: The student drawing represents elements within their sensory plane, generally as effects or consequences of the earthquake phenomenon, recognizable as movements of the ground's surface or results of the movement. The information derived from the representation was insufficient for the researchers to interpret an explanation beyond the child's perceptible plane, for instance, based on non-visible entities. These drawings frequently have a baseline to delimit the ground line (in a continuum, backstitch, or oblique) or function as object support. Some graphics also wrote words related to "movement" or "seism," etc., while others designed zigzag or wavy lines to represent the consequences of movement on the objects, as Figure 3 shows. Thus, we interpreted these drawings as a sensory level of reasoning because the cognitive operation is based on entities or

**Figure 3.** Examples of level 1. Drawing (**A**) illustrates a field with plants moving, (**B**) a ground line with scared children moving, and a happy face below the baseline. **Figure 3.** Examples of level 1. Drawing (**A**) illustrates a field with plants moving, (**B**) a ground line with scared children moving, and a happy face below the baseline.

Level 2: Some representations or elements are beyond the students' primary sensory or perceptual level. The drawings in this category (Figure 4, in which we have translated what the students wrote in their drawings) usually present changing elements, for instance, beneath the ground, or views from outside planet Earth, commonly represented by a baseline–ground line or object support–to express a division between the elements perceived by children and the elements not perceived but conceptualized and represented as the possible causes of earthquakes. In these types of drawings, we observed an attempt at expressing a causal relationship between the consequences of the earthquake (i.e., beyond the baseline) and their origin (i.e., beneath the baseline); however, it is not evident that these changing entities are related to the interactive basis of TPT, such as movement, friction or a crash of plates, or the dynamics of the internal structure of the Earth. Thus, we interpret a more complex level of reasoning than in level 1 because children are reasoning through elements or processes that are further from their immediate sensory experience and trying to express causal thinking, nonetheless not yet at a level that uses the parts of the theory to represent a causal process or ongoing mechanism. Level 2: Some representations or elements are beyond the students' primary sensory or perceptual level. The drawings in this category (Figure 4, in which we have translated what the students wrote in their drawings) usually present changing elements, for instance, beneath the ground, or views from outside planet Earth, commonly represented by a baseline–ground line or object support–to express a division between the elements perceived by children and the elements not perceived but conceptualized and represented as the possible causes of earthquakes. In these types of drawings, we observed an attempt at expressing a causal relationship between the consequences of the earthquake (i.e., beyond the baseline) and their origin (i.e., beneath the baseline); however, it is not evident that these changing entities are related to the interactive basis of TPT, such as movement, friction or a crash of plates, or the dynamics of the internal structure of the Earth. Thus, we interpret a more complex level of reasoning than in level 1 because children are reasoning through elements or processes that are further from their immediate sensory experience and trying to express causal thinking, nonetheless not yet at a level that uses the parts of the theory to represent a causal process or ongoing mechanism. *Educ. Sci.* **2021**, *11*, x FOR PEER REVIEW 12 of 20

**Figure 4.** Examples of level 2. Drawing (**A**) represents moving buildings on the surface and Earth layers beneath, (**B**) shows a broken building and elements under the base line labeled as "Plates". **Figure 4.** Examples of level 2. Drawing (**A**) represents moving buildings on the surface and Earth layers beneath, (**B**) shows a broken building and elements under the base line labeled as "Plates".

nents are interacting, changing position, or moving. These concepts are expressed as a causal explanation of the earthquake, directly connected with TPT, such as movement, friction or a crash between plates, or the Earth's internal structure dynamics. We observed drawings that included the causes and consequences of the phenomenon, usually with arrows or labels indicating the name of the components (i.e., epicenter, interaction, etc., illustrated in Figure 5) or the direction of the movement. Thus, we interpret these drawings as precursor models used by the participants to express a causal relationship between the phenomena and the underpinning theory, which means a qualitative leap of children's reasoning towards thinking with non-visible theories to explain processes or ongoing mechanisms. It is worth noting that we made no judgment of the conceptual accuracy

Level 3: The representations in this level were more complex in comparison with level 2. The drawings include elements outside or beyond the children's primary sensory

Conceptual accuracy refers to the degree of content correctness in the scientific use of concepts, terms, or postulates in the drawing. Although in other works with secondary students' explanation, a conceptual inaccuracy in the written explanation implies coding in level 0, e.g., [44], in this study, we consider that primary students can have inaccuracies expected because they had only started to learn about the content. Thus, we decided to give value even to explanations that were not totally precise but showed the advance in the reasoning process. For instance, in Figure 5, student's drawing A represented the causes of earthquakes under the baseline, reasoning with abstract entities, represented a model of interaction, signaling a black point where the energy releases as "the epicenter." Although the correct term should be "the hypocenter," we made no judgment of the conceptual accuracy in the representation and consider it is an advance in the expressed rea-

soning regarding levels 1 or 2. Thus, we categorized it at level 3.

presented through the representation.

Level 3: The representations in this level were more complex in comparison with level 2. The drawings include elements outside or beyond the children's primary sensory or perceptual level. However, the difference with level 2 is that, in level 3, these components are interacting, changing position, or moving. These concepts are expressed as a causal explanation of the earthquake, directly connected with TPT, such as movement, friction or a crash between plates, or the Earth's internal structure dynamics. We observed drawings that included the causes and consequences of the phenomenon, usually with arrows or labels indicating the name of the components (i.e., epicenter, interaction, etc., illustrated in Figure 5) or the direction of the movement. Thus, we interpret these drawings as precursor models used by the participants to express a causal relationship between the phenomena and the underpinning theory, which means a qualitative leap of children's reasoning towards thinking with non-visible theories to explain processes or ongoing mechanisms. It is worth noting that we made no judgment of the conceptual accuracy presented through the representation. *Educ. Sci.* **2021**, *11*, x FOR PEER REVIEW 13 of 20

**Figure 5.** Examples of level 3. (**A**) represents a damaged house, a sad person on the surface and under the herb line a point of interaction labelled "the epicenter" with facing arrows. (**B**) shows trees moving, a scared person above the baseline and two blocks moving labelled as "Tectonic plates are moving" under the baseline. **Figure 5.** Examples of level 3. (**A**) represents a damaged house, a sad person on the surface and under the herb line a point of interaction labelled "the epicenter" with facing arrows. (**B**) shows trees moving, a scared person above the baseline and two blocks moving labelled as "Tectonic plates are moving" under the baseline.

*3.4. Boundaries for Interpreting "Qualitative Leaps" to a Superior Level*  Our research found three qualitative leaps of expressed reasoning in students through drawn explanations, which help us interpret a hypothetical progression of reasoning. The first one (a) marks the level at which we can affirm interpretable reasoning about the phenomenon. The second one (b) refers to an advance from the upper anchor of level 1 to the inferior anchor of level 2. The third leap (c) occurs between the upper anchor of level 2 to the low anchor of level 3. (a) The entry point to the hypothetical progression of reasoning is the connection of the explanation with the phenomenon of interest. In this case, we observed the leap between level 0 and level 1 when the students represented the effects or consequences of earthquakes. In addition, they recognized that, in the context of learning about the internal dynamics of the Earth, the cognitive task that required drawing "why does Conceptual accuracy refers to the degree of content correctness in the scientific use of concepts, terms, or postulates in the drawing. Although in other works with secondary students' explanation, a conceptual inaccuracy in the written explanation implies coding in level 0, e.g., [44], in this study, we consider that primary students can have inaccuracies expected because they had only started to learn about the content. Thus, we decided to give value even to explanations that were not totally precise but showed the advance in the reasoning process. For instance, in Figure 5, student's drawing A represented the causes of earthquakes under the baseline, reasoning with abstract entities, represented a model of interaction, signaling a black point where the energy releases as "the epicenter." Although the correct term should be "the hypocenter," we made no judgment of the conceptual accuracy in the representation and consider it is an advance in the expressed reasoning regarding levels 1 or 2. Thus, we categorized it at level 3.

#### the ground move?" involves a specific phenomenon—an earthquake. Level 1 is mi-*3.4. Boundaries for Interpreting "Qualitative Leaps" to a Superior Level*

nor complex because the student only needs to identify a logical connection within the task's context. For example, in Figure 2B, the planet Earth drawing was categorized at level 0, missing data. However, in Figure 3A, at the bottom anchor of level 1, we considered the black lines around the plants in the soil to represent movement, according to the categories by Tang et al. [55], which signal a consequence of the Our research found three qualitative leaps of expressed reasoning in students through drawn explanations, which help us interpret a hypothetical progression of reasoning. The first one (a) marks the level at which we can affirm interpretable reasoning about the phenomenon. The second one (b) refers to an advance from the upper anchor of level 1 to

(b) Comparing the upper anchor of level 1 to the inferior anchor of level 2, we can ob-

(c) Between the upper anchor of level 2 and the inferior anchor of level 3, we interpret a leap signaled by some representation elements connecting with the modeling process in science education, in the labeling of "Plates" in Figure 4B. However, no interaction between the components was expressed. The sophistication was demonstrated by the more explicit representation of the interaction between unobservable entities. Figure 5A and 5B represents cause, consequences, and activities between the components of

TPT. They show reasoning with theory to explain a natural phenomenon.

non-perceptual plane, beneath this line, there are no recognizable elements. On the contrary, in Figure 4A, it is possible to observe the same ground line but with a representation of the Earth's layers similar to the static model. Thus, we interpreted the increased complexity of the child's recognition of possible causes of the phenomenon

with an incipient link to the TPT.

earthquake.

the inferior anchor of level 2. The third leap (c) occurs between the upper anchor of level 2 to the low anchor of level 3.


#### **4. Discussion**

In this study, we sought to explore the characteristics of students' expressed reasoning through drawn explanations in the context of learning about earthquakes at an early stage of formal instruction. We developed an instrument based on previous research to elicit and analyze fourth graders' scientific reasoning based on theory through their drawn explanations. The analysis allowed the recognition of three levels of scientific reasoning, which were possible to characterize in the participants of this study. Consequently, our findings answer the question proposed: What characterizes students' expressed reasoning in drawn explanations in the context of learning about earthquakes?

In summary, at level 0, topic-specific reasoning was not interpretable from the representational explanation. In contrast, at level 1, students' reasoning was based mainly on the perceptible entities associated with the consequences of the phenomena. Drawings characterized as level 2 showed that children's reasoning starts to connect some theory elements as a first attempt to explain the causes of a phenomenon. Despite this, levels 1 and 2 lack causal relations using the theory. Finally, in level 3, students could express their scientific reasoning about the phenomenon by linking elements of TPT to explain the causes and effects of a phenomenon as a precursor model, considered as cognitive schemata compatible with scientifically appropriate knowledge [27].

Moreover, our study found qualitative leaps between the children's levels of expressed scientific reasoning focused on the connection with the phenomena under investigation, the emergence of the divisions of the perceptual and non-perceptual plane, and the presence of recognizable elements of the theory as part of the representation of the explanation. Some of the more advanced features expressed by the participants in our study presented similar characteristics to those of Perkins and Grotzer [40]. Specifically, we interpreted sophistication from static comprehension to an interactive activity between the non-observable or theoretical entities.

According to Yao and Guo [44], the students first relate indirectly to facts and theory through models before their scientific reasoning evolves, progressively approaching scientifically accepted models. Analyzing students' drawings as an expression of their reasoning process gave us evidence for interpreting more sophisticated reasoning during the learning unit and students' drawings as precursors of scientific models. This idea might be construed from a transformation of embedded intuitive theories through language [18] and deliberate thinking [19]. Moreover, our study expanded the literature to other forms of capturing advances of students' reasoning through their creative activity of drawing explanations, which represents a complement to the current instruments to analyze students' written explanations [31,43,44].

However, simultaneously analyzing and fostering students' explanation construction based on theories, principles, or concepts is still a challenge at the early stages of formal learning [6]. Given this, we need to understand that students are still constructing the meaning of the scientific concepts involved when explaining. In primary education, they also develop essential skills such as explaining for scientific purposes or using models to explain the world. Thus, we emphasize the importance of supporting students to build these capacities and not underestimating their possible ability to express their scientific reasoning and knowledge through formats more familiar to them, such as drawings. Combining forms for approaching scientific reasoning and learning might mean a synergistic effort to scaffold the emergence and sophistication of reasoning, the conceptual understanding of children, and the development of essential skills. Our results resonate with prior research showing the need to combine diverse data sources to interpret children's scientific learning [27].

Park et al. [50] discussed pictorial representation as a complementary format to explore students' ideas. In this, they argued that drawings involve implicit information that is connected to other external representations. Indeed, the ways students express themselves about a concept or idea might be different when they do it verbally and pictorially, or exclusively verbally. We believe that for younger students, it is through drawings or representations that they are building scientific ideas and connecting them to other types of representations. We know that for students to construct scientific concepts, multimodal languages support processes related to sensemaking, scientific explanation construction, and scientific concept development [21]. It implies that employing exclusively visual or verbal representation during teaching might limit students' learning process. Considering pictorial representations as part of multimodal language supports students in building concepts that are vehicles for expressing their reasoning. By having students use verbal communication only for concept construction, incorporating pictorial representations might result in more prosperous, more robust, and connected ideas for concepts construction, perhaps involving a re-conceptualization due to changing modalities. This is because constructing explanations seen from a sociocultural perspective is a knowledge integration learning artifact, in which the students connect what is already known with their experiences and conceptual elements to give scientific support for certain phenomena [25].

The instrument used to analyze scientific reasoning based on theory for primary students was demonstrated to be sensitive enough to detect the sophistication of these elements of reasoning during a learning unit of eighteen hours in the context of this study. Specifically, we observed an increase in level 1 and 3 categorized drawings between the learning sequence's beginning—S1—and the end—S2 (Figure 1). Thus, we can conclude that, after participating in a formal instruction process, some participants in our study could express more sophisticated reasoning with a causal link in their drawn explanations. We agree that explaining phenomena provides an optimal scenario to connect students with socio-scientific issues [30,31], and our study adds that student drawings can be a source of expressed reasoning and, at the same time, a learning activity that activates and allows enacting or triggering of specific systems of reasoning.

Nonetheless, the instrument allowed the identification of a significant group of participants who did not show evidence of operating cognitively with unobservable entities to connect the causes and consequences of the phenomena under study. After the learning unit was finished, this gap was observed, with students immersed in a high-risk context, adding familiarity with the phenomenon. We recommend providing opportunities to learn to link phenomena and their causes in this and several other topics and conducting more research to determine the obstacles to student advancement in reasoning levels. Still, we observed instances of expressed reasoning regarding context-related situations before formal learning started at school. The entry point to the hypothetical reasoning progression was the connection of the explanation with the studied phenomenon. This finding coincides with studies that show the starting point for explanation-construction is the phenomenon [4], which helps to afford the need to generate explanations. By fourth grade, Chilean students have likely already had some daily life experiences with earthquakes and can nurture their reasoning process about the environment in which they live. Thus, the fact that our study considered the early stages of formal learning and identified what ideas the students had already formed in their representations for constructing explanations is valuable. Further research could illuminate the role of local context in early scientific reasoning levels, not only on how scientific reasoning about earthquakes develops throughout the school trajectory but also extending the use of such instruments to other subjects, areas, or demanding tasks.

This study has some limitations. First, using a strategy designed in a different context and language might cause cross-cultural issues. We adapted the frame suggested by Park et al. [50] according to the context of the study. Still, we also acknowledge the particularities of Chile as having a high risk of disaster (e.g., earthquakes). Thus, the learning approach to these phenomena may vary from those whose context does not include risk or whose geographical reality is very different. However, this point also represents a possible subject for future researchers to explore: the extent to which proximity to a phenomenon might imply a variation in the way students think about it.

Additionally, some elements of students' drawn explanations went beyond the frames of our analysis, for instance perspectives from outside the planet that combined astronomical concepts. Although we treated those features as exceptions in our study, perhaps representing a limitation, we believe a second perspective on these types of data is crucial to challenge adults' beliefs about the abstraction capacity of children and the way they visualize phenomena and their causes. Moreover, we recognize our study has a small sample size for going beyond descriptive analysis. Thus, we encourage further research to work with larger groups of students for complementary validation purposes.

Regarding the validation of the rubric, in this study, we went through a content validation through a panel of experts and a small pilot study before analyzing the data sets. Due to the small sample size and the study's exploratory nature, we could not run factor analysis or more sophisticated processes, strengthening the significance and or generalizability of the results.

Nonetheless, we consider this study as a first approach interpreting primary students' reasoning in phenomena explained by theory, with an educational significance in the field of science education. Other researchers might take the advances of our work and, for instance, compare pre–post drawings in specific groups of students, or use a repeated-measurements design focusing on learning the topic or conceptualization of the phenomena. Hence, we suggest future research gathering evidence of the leaps shown in our study but exploring them in the light of learning progressions of individual students. This exploration might complement the current results to emphasize the connections between understanding phenomena, theories, or concepts and learning, to establish learning trajectories in science education.

#### **5. Conclusions**

The current study allowed us to characterize students' scientific reasoning through drawn explanations. We presented a helpful instrument to identify cognitive leaps between concrete expressed reasoning levels and more abstract ones, including causal links between phenomena and theory. It is a methodological innovation to approach young students' learning and reasoning development from an interdisciplinary perspective that combines education and cognitive science. Our research explicitly links science learning and cognition by highlighting and approaching children's reasoning about natural phenomena underpinned by theory. This development expands the current instruments available to notice the complexity of scientific reasoning of young children when they are at the first moments of learning models, theories, or abstract postulates that sustain the causes of phenomena. There is a methodological advance considering that most of the current instruments relate to explanations based on evidence in written formats.

In applying the developed rubric, we observed sophistication in students' scientific reasoning when provided a formal learning opportunity, resulting in some students progressively connecting their ideas to a scientific theory. Our study allowed exploration of students' progressive development of the causal reasoning required to construct explanations. Constructing explanations based on theory from primary school is a relevant teaching and learning practice to develop at an early stage of learning, considering that secondary and college students have limitations to using their scientific knowledge to establish causal links when they construct explanations. Furthermore, identifying scientific reasoning levels at the early stages of learning allows conceptualization of scientific reasoning as a trajectory. Thus, we can observe more precisely where students begin this form of complex knowledge and how it will eventually progress. By identifying and understanding this trajectory and the qualitative leaps, teachers, educators, and researchers can better scaffold the learning process and the development of context-related scientific reasoning, providing opportunities to support this development promptly. The detailed description of these findings helps researchers interested in this field adapt, reframe, and test in different ways the analysis we have done, allowing projection of transference of the interpreted reasoning of the rubric of this research to other topics. It would make the progressive approach to thinking in different disciplines visible and promote students' reasoning in the school. This idea resonates with theoretical frameworks used for understanding of the construction of explanations as epistemic processes, which broadens the interest of this article to other areas beyond the content of the application in our study.

Teachers' support of children's reasoning in the classroom might take the form of distributed scaffolding. For instance, giving prompts with initial questions such as in the present study "why do you think this phenomenon happens", and moving forward to students to revise and enrich their initial explanations during the learning of the content advance. The scaffolding seeks to transfer the responsibility gradually to the student, promoting autonomy. In primary education, where students are diverse in autonomy degrees, generating group discussions about explanations is an option, considering that science practices also imply peer-reviewing ideas and claims to compare and contrast to evaluate their scope and limitations. This strategy also connects with positioning science construction as a collective activity, introducing children to elements of Nature of Science. We strongly believe that classroom activities oriented to develop students' reasoning processes should encourage students to express their ideas in diverse formats, such as the causes of phenomena. Then, linking those with the scientific support through concepts, theories, or postulates that are usually more abstract entities to reason. However, this approach needs teachers to consider that students' common sense is part of their implicit theories that allow them to make sense of emergent phenomena, thus relevant for transformation and not represent merely knowledge to discard during science lessons. We know that teachers tend to suppress ideas that might look wrong as they are expressed in more traditional science classrooms. Still, we want to stress that responsive science teaching gives value to the students' existing ways of thinking to construct new understanding, further develop

their reasoning into a more scientific one, managing the supports strategically that students need promptly.

Furthermore, our work supports understanding primary students' reasoning considering current educational challenges, affording students' thinking processes through authentic practices, such as constructing explanations based on context-related phenomena. Moreover, we see the explanation-construction of relevant phenomena as a participatory action for responsible citizenship that can be implemented in primary education to promote high-leverage practices such as explaining and modeling, as was mentioned in our theoretical framework. Thus, we highlight that, even at the early stages of formal science learning, students can transform their ideas into expressions of context-related reasoning, for instance, through drawings that act as learning samples of explanations represented at the first stages of their models to explain natural phenomena. This fact emphasizes the importance of recognizing young children as active constructors of knowledge, showing that some can go beyond their immediate experience to logically link a phenomenon with its underpinning theory. Constructing explanations about world phenomena and expressing students' reasoning in formats aligned with their action, drawing creative activity is a more abstract and complex process worthy of considering by researchers, educators, and teachers interested in the multidisciplinary innovations for understanding learning processes and outcomes.

**Author Contributions:** Conceptualization, V.M.C. and P.M.M.; methodology, V.M.C. and P.M.M.; formal analysis, V.M.C., P.M.M. and P.G.M.; writing—original draft preparation, V.M.C. and P.M.M.; writing—review and editing, P.G.M.; project administration, V.M.C.; funding acquisition, V.M.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** CONICYT/FONDECYT/11181050, currently Agencia Nacional de Investigación y Desarrollo ANID/FONDECYT/ 11181050.

**Institutional Review Board Statement:** This study was approved by the ethics committee of the Pontificia Universidad Católica de Chile, under number 180514006.

**Informed Consent Statement:** All participating children had active parental consent, meaning that parents were informed and agreed to their child's participation in the study.

**Data Availability Statement:** Data are available if required.

**Acknowledgments:** We thank the article's anonymous reviewers for their thoughtful comments.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the study's design, in the collection, analyses, or interpretation of data, in the writing of the manuscript, or in the decision to publish the results.

#### **References**

