Introduction

Scientific reasoning (SR) is defined as the abilities and cognitive processes required to use the scientific method (Díaz et al., 2021). SR helps us understand connections between actions and outcomes in daily life (Kuhn, 2012), enables evidence-based decisions in our professions (Berndt et al., 2021), and enhances students’ understanding of science subjects (Kuhn, 2011). Well-developed SR is characterized by the systematic, intentional obtainment of relevant information and knowledge, complemented by reflective abilities (Morris et al., 2012).

Research on SR focuses on the three domain-general abilities of hypothesis generation, investigation and experimentation, and evidence evaluation (Zimmerman & Klahr, 2018). Albeit studies report indications of SR abilities in children, adolescents and adults often fail on SR tasks (e.g., Kuhn, 2007, 2012). For example, they misconstrue or dismiss contradictory evidence, engage in unsystematic and uninformative experimentation, and seek confirmatory evidence for their prior beliefs (see Zimmerman & Klahr, 2018).

The literature contains two explanatory accounts for this seeming contradiction. The first stems from the literature describing the development of scientific reasoning (DSR). It argues that findings exaggerate young children’s SR abilities, highlighting that corresponding tasks’ demands are considerably lowered, nevertheless leading to children making errors in exploration, experimentation, and evidence evaluation (e.g., Kuhn, 2011). It also notes that children tend to favor effect-producing strategies over informative interventions (i.e., the “Positive Testing Strategy”; Shtulman & Walker, 2020), which indicates a lack of systematicity, intentionality, and metaconceptual understanding. The second explanatory account derives from the causal learning literature. Building on ample research, it maintains that “children’s learning mechanisms do indeed resemble the basic inductive processes of science” (Gopnik, 2012, p. 1624), and posits that previous studies have misinterpreted adolescents’ and adults’ valid inquiry and evidence evaluation strategies in SR tasks as uninformative and false (Lapidow & Walker, 2021).

The present essay first summarizes the two accounts’ interpretations of children’s hypothesis generation, investigation and experimentation, and evidence evaluation. In conjunction, their core claims regarding the SR failures of adolescents and adults are reviewed, and implications for education are derived. After that, the accounts are compared and discussed in their explanatory power for the seeming contradiction in the literature. In closing, limitations and future directions are outlined.

DSR Literature

The DSR literature notes that children often struggle to differentiate between theories, hypotheses, and evidence, cling to false prior beliefs in the face of contradictory evidence, and overlook confounding variables when determining causal effects (e.g., Kuhn, 2011; Schauble et al., 1991; Tschirgi, 1980; Zimmerman & Klahr, 2018). Moreover, to engage in SR tasks, children require lowered task demands and guiding constraints that are unrepresentative of the real world (e.g., forced-choice paradigms with just one causal variable; Kuhn, 2012; Zimmerman & Klahr, 2018). Leaning on these results, DSR scholars posit that children show shortcomings in their investigation, experimentation, and evidence evaluation in SR tasks (e.g., Kuhn, 2012). Notably, evidence around children’s hypothesis generation skills is still scarce. One study by Piekny and Maehler (2013) reported that preschoolers show difficulties completing related tasks, possibly due to overwhelming metacognitive requirements.

For the cases where children succeed on rudimentary SR tasks, DSR scholars stress that they do so without the intentionality, systematicity, and meta-cognitive skills characteristic of well-developed SR (e.g., Kuhn, 2012). For example, Tschirgi (1980) and Schauble et al. (1991) concluded from their studies that children prefer to produce interesting outcomes over deconstructing cause-and-effect relationships. This points toward children lacking an explicit understanding of what makes a good experiment and effective strategies for obtaining informative evidence. Notably, only few empirical studies on DSR have explicitly targeted metacognitive awareness and intentionality in SR tasks. Among them, interestingly, is contradictory new evidence by Köksal-Tuncer and Sodian (2018) and Köksal et al. (2021): They found preschool-aged children can differentiate between uninformative and informative evidence, judge their own knowledge states (e.g., ignorance or lack of evidence to make a claim), and provide verbal disconfirmation when presented with false causal claims.

The broader claim by the DSR literature remains that children and adults primarily differ in their metacognitive abilities to adequately assess and use evidence to revise false prior beliefs and differentiate between theory and evidence (Zimmerman & Klahr, 2018). Albeit it is acknowledged that adults usually have well developed metacognitive abilities, they still often go on to fail on SR tasks (e.g., Kuhn, 2007). The DSR literature explains that adult SR tasks often build on prior knowledge and entail more complex causal structures (e.g., multiple potential causal variables and interactions). If adults’ prior knowledge and beliefs conflict with the evidence in the task, they have to reconcile the two – a difficult task that requires prior skill-building (Kuhn, 2012). It follows that we should neither expect untrained adults to perform well on SR tasks nor children to become adults with well-developed SR without training and instruction (Shtulman & Walker, 2020).

Hence, schools need to devote attention to the explicit training of SR abilities and related metacognitive skills. For one, they can implement question-asking practices in their classrooms, such as encouraging students to ask “why?” and “what if?” questions. This can help them identify what they do not yet know, support their explanations, and build a habit of considering alternative hypotheses (Shtulman & Walker, 2020).

Causal Learning Literature

The causal learning literature has termed children “little scientists” not only due to their intuitive theories about the world (e.g., intuitive physics), but because their “learning mechanisms do indeed resemble the basic inductive processes of science” (Gopnik, 2012, p. 1624). Starting in preschool age, they can design interventions to learn about causal structures, systematically test hypotheses, engage in more exploration if evidence is confounded, choose informative interventions over uninformative ones, and learn from these processes to revise their hypotheses and theories (Köksal-Tuncer & Sodian, 2018; Lapidow & Walker, 2020; McCormack et al., 2016; L. E. Schulz & Bonawitz, 2007). The causal learning literature interprets these findings as evidence of children’s investigation, experimentation, and evidence evaluation skills being akin to those of scientists (Lapidow & Walker, 2021).

Notably, children’s use of evidence and SR strategies alone is insufficient to infer intentionality and conceptual understanding (Köksal-Tuncer & Sodian, 2018). Some recent causal learning studies on exploratory play indicate that children are aware of different types of evidence to some extent. For example, they actively isolate potential causal variables when evidence is ambiguous (Cook et al., 2011). However, research explicitly targeting metacognitive awareness and conceptual understanding is thin spread.

Notably, the causal learning literature documents a trend of children’s SR abilities steadily increasing with age: “Even the youngest children tested appear to grasp already the fundamental concepts and skills of scientific inquiry and inference, and this is followed by steady, incremental progress towards consistently correct performance with increasing age.” (Lapidow & Walker, 2021, p. 4). This is backed by recent longitudinal studies showing that albeit SR performance does not improve linearly, older children generally do show growth in their SR abilities (e.g., in experimentation and evidence evaluation; Lazonder et al., 2021; Piekny & Maehler, 2013).

Lapidow and Walker (2021) posit that especially adults’ attempts at SR tasks may have been misinterpreted as false and irrational despite being valid. In their “interventionist theory of causal learning,” causal knowledge entails knowing if throwing a stone (X) into a window causes it to break (Y). In contrast to the DSR literature’s disapproval of PTS, Lapidow and Walker argue that PTS is a valid choice for determining causality: Collecting data on one instance where X is absent cannot determine with certainty what happens if X is in play. In contrast, repeatedly testing what happens if X is present can – it addresses the question of whether the outcome (a broken window) invariantly occurs when the cause (throwing a stone into the window) is in play. Therefore, children’s and adults’ preference for PTS should not be seen as an erroneous, uninformative strategy but “as a rational and informative approach to generating evidence” (Lapidow & Walker, 2021, p. 7). Hence, the authors note that evidence found for children’s and adults’ failures on SR tasks may well be less severe than the DSR literature argues.

As for implications for education, the causal learning literature notes that children learn considerably from their exploration and play. Thus, schools and early childhood education should refrain from structuring and formalizing their environments to mirror formal educational contexts (Gopnik, 2012). Instead, they should maintain room for children to learn informally, e.g., through pretend play. While explicit instruction is not advisable, educators and caretakers can, for instance provide scientific explanations related to everyday events or carefully introduce children to counterintuitive information and ask them to explain their intuitions (Gopnik, 2012).

Comparison of Accounts

For three reasons, I conclude that the causal learning literature has more been consistent with recent findings and provides a more convincing explanatory account. First, there is now ample evidence of children’s abilities around investigation, experimentation, and evidence evaluation in SR (e.g., Köksal-Tuncer & Sodian, 2018; Lapidow & Walker, 2020; McCormack et al., 2016; Schulz & Bonawitz, 2007). The DSR literature traditionally argued that children are fundamentally deficient in these areas but has come to concede that children do at least show certain SR precursors (e.g., Shtulman & Walker, 2020). The documented trend indicates that children’s SR abilities develop with increasing age, underlined by longitudinal studies (Lazonder et al., 2021; Piekny & Maehler, 2013), aligning with the causal learning literature’s claims (e.g., Lapidow & Walker, 2021). Relatedly, the DSR literature argues that children and adults differ fundamentally in their metacognitive abilities to revise their beliefs and differentiate hypotheses and evidence (e.g., Zimmerman & Klahr, 2018). However, new research now finds that young children can indeed distinguish between hypotheses, theories, and evidence (e.g., Köksal-Tuncer & Sodian, 2018a). Despite adults allegedly being more well-developed metacognitively, they still often fail on SR tasks (Kuhn, 2007; Lapidow & Walker, 2021). The DSR literature has argued that the average adult lacks the explicit training and education to navigate researchers’ demanding SR tasks (Shtulman & Walker, 2020), but the underlying prediction was nevertheless rendered false.

Lastly, the DSR literature has stressed that children often engage in SR tasks with a “Positive Testing Strategy” (PTS; Shtulman & Walker, 2020). A considerable part of this claim rests on early studies (e.g., Schauble et al., 1991; Tschirgi, 1980) that remain widely cited in current theoretical work and reviews (Lapidow & Walker, 2021; Zimmerman & Klahr, 2018). However, recent findings from Lapidow and Walker (2020) indicate that their preference for desirable outcomes does not primarily drive children. Moreover, Lapidow and Walker (2021) have posited a theoretical account explaining how PTS can be considered an effective and informative hypothesis testing approach instead of an erroneous one.

Limitations and Conclusion

Scientific reasoning (SR) entails the abilities and cognitive processes required to use the scientific method, specifically hypothesis generation, investigation and experimentation, and evidence evaluation (Díaz et al., 2021). Literature around SR has reported that young children engage in SR similar to scientists, but that adolescents and adults often fail on SR tasks (e.g., Zimmerman & Klahr, 2018).

The causal learning literature has provided a convincing explanatory account for this seeming contradiction through its extensive documentation of children’s early SR abilities (e.g., Gopnik, 2012), first reports of children’s awareness and systematicity in SR (e.g., Cook et al., 2011; Schulz & Bonawitz, 2007), the supported prediction of continuous SR development (e.g., Lazonder et al., 2021; Piekny & Maehler, 2013), and a novel theoretical account identifying PTS as an informative and valid approach to SR (Lapidow & Walker, 2021).

It should be noted that drawing a sharp distinction between the causal learning and DSR literature may no longer be appropriate: Recent longitudinal studies and works that aim to tackle the gaps around metacognitive skills of children (e.g., Köksal-Tuncer & Sodian, 2018a; Lapidow & Walker, 2020; Lazonder et al., 2021) effectively build on both literature strands, and widely cited scholars such as Schulz produce findings in cognitive development as well as causal learning (e.g., Schulz, 2012; Schulz & Bonawitz, 2007). A different framework for comparing opposing accounts in the literature should be considered for future reviews.

Ultimately, scholars from both strands have contributed much to our understanding of children’s development in SR and the mechanisms of their exploratory play. First educational policy suggestions for nurturing these aspects have been made (e.g., Gopnik, 2012). If early childhood education and schools build on these findings, society will benefit from more citizens well-versed in SR, for example, holding a better understanding of science subjects (Kuhn, 2011), and making better evidence-based decisions in their professions (e.g., Berndt et al., 2021).