Identifying precursory concepts in evolution during early childhood – a systematic literature review

ABSTRACT

Difficulties in understanding evolution are often rooted in early childhood, arising from naïve assumptions and cognitive biases. However, literature reviews mainly focus on school and university students’ understanding of evolution, with only limited comprehensive reviews on children in early childhood aged up to 7 years. This systematic review aims to capture precursory concepts in evolution and influencing cognitive biases as documented in the empirical literature. Searches of three databases identified 204 articles, of which 26 were used for further analyses after screening for eligibility. The analyses revealed that even young children are capable of understanding the basic mechanisms of core concepts in evolution, such as variation, inheritance, and natural selection. However, while children’s understanding of the inheritance concept has been investigated intensively, their understanding of variation lacks in-depth research despite its probable influence on natural selection. Existing evidence is contradictory concerning the usefulness of children’s cognitive biases for learning core concepts in evolution: These can serve as stepping stones for learning evolutionary principles, but their usefulness is questioned if children have already developed scientifically correct explanations. More research is clearly needed concerning the reciprocal effects of children’s precursory core concepts in evolution in order to develop effective learning interventions for children.

KEYWORDS:

• Early childhood education
• variation
• inheritance
• natural selection
• cognitive biases

Introduction

Despite the central position of biological evolution within the life sciences (Mayr, 1982), the theory of evolution often clashes with perceived common sense – a frequent cause of resistance against learning scientific ideas (Bloom & Weisberg, 2007). A large body of empirical research indicates that grasping evolution is troublesome for learners at each phase of their educational career (e.g. Banet & Ayuso, 2003; Bishop & Anderson, 1990; Crawford et al., 2005; Nehm & Reilly, 2007; Harms & Reiss, 2019; Kampourakis & Zogza, 2008; Lawson & Thompson, 1988). For example, Kampourakis and Zogza (2008) found that secondary school students often explain evolution through final causes rather than through evolutionary mechanisms. Moreover, Nehm and Reilly (2007) found that in their study only 3% of first-year biology majors used a multiple component concept of natural selection, while alternative conceptualisations have been used by pre-service (Crawford et al., 2005) and in-service science teachers (Nehm & Schonfeld, 2007).

A review of articles published from 2000 to 2014 (Glaze & Goldston, 2015) regarded the acceptance or rejection of the theory of evolution as the opposite ends of a continuum that is influenced by an individual’s concepts formed long before formal schooling. While that review focused particularly on secondary and teacher education (Glaze & Goldston, 2015), the review of Russell and McGuigan (2015) provides a bibliography ‘designed to develop practical guidance for teaching Evolution and Inheritance at KS1 and KS2’ (p. 1). In their review, they extended the focus from evolution as a single concept to neighbouring concepts such as inheritance. Despite this growing evidence, a systematic review of research on children’s precursory concepts in evolution at the end of early childhood is still missing. The fact that research about children’s understanding of evolution in early childhood has been neglected for a long time is mirrored by the statement of Shtulman and Harrington (2016) concluding that only ‘two studies have explored whether children can be taught evolutionary ideas at all’ (p. 1224).

Cognitive development between the ages of 2 and 7 has long been referred to as the preoperational stage in which children are not able to (1) perform mental operations such as classification by joining things into one class, or (2) provide causal explanations that go beyond intentional causality (Piaget, 1964). However, more recent research on cognitive development in early childhood has revealed that even before the age of 7 children have naïve theories on the biological, physical, and social world – coined as naïve biology, physics, and psychology – because they provide ontological distinctions and causal explanations (Bloom & Weisberg, 2007; Hatano & Inagaki, 1994; Wellman & Gelman, 1992). For example, Springer and Keil (1989) were amongst the first to show that children aged 4 through 7 years consider features as inheritable if these features perform a biological function for animals, whereas features with social or psychological functions are not considered as inheritable. Thus, young children seem to have a notion of biological inheritance. However, children are in danger of developing resistance to science when scientific concepts are contrary to their intuitive assumptions (Bloom & Weisberg, 2007). Thus, difficulties in understanding evolution are often rooted in early childhood, arising from intuitive assumptions and cognitive biases such as ‘promiscuous’ teleology (Kelemen, 1999, p. 1441, 2012), and can persist into adulthood (Bloom & Weisberg, 2007). In fact, at the age of 7 through 8 years, those early cognitive biases can reinforce each other (e.g. teleological beliefs can enhance essentialist tendencies; Emmons & Kelemen, 2015). Research has also revealed that both scientific and intuitive concepts of evolution coexist within a learner (Shtulman & Harrington, 2016) and that the presence and absence of specific cognitive biases within a learner are context-dependent (Nehm & Ha, 2011; Opfer et al., 2012). Despite cognitive science and science education research on how children’s precursory concepts and cognitive biases develop as naïve theories and shape their learning in later stages of education after the age of 7 (Emmons & Kelemen, 2015; Tytler & Prain, 2009; Vosniadou & Brewer, 1992; Vosniadou et al., 2001), only few curricular or pedagogical measures engage with such biases (Opfer et al., 2012). Based on the International Standard Classification of Education (ISCED; UNESCO Institute for Statistics, 2012), the age of 7 marks the latest age for the transition from early childhood education (ISCED 0) to primary education (ISCED 1) so that pedagogical measures should account for children’s precursory concepts developed before the age of 7. Therefore, in our literature review, we: (1) consider the question that intuitive concepts can be useful stepping stones in early childhood for learning evolution; (2) aim to capture the relevant precursory core concepts of evolution documented in the literature; and (3) determine which precursory concepts serve as stepping stones and which educational interventions may prepare children to understand the theory of evolution before the age of 7.

Theoretical background

Core concepts to explain evolutionary change

The theory of evolution is the central and overarching theme that unites the life sciences. Despite a growing understanding of the diverse processes involved in evolutionary change, such as genetic drift, genetic linkage, and endosymbiosis, many evolutionary biologists agree that the major mechanism for explaining evolutionary change is the process of natural selection (Endler, 1986; Mayr, 2001). The construct of natural selection is generally defined in terms of essential elements or key concepts (e.g. Anderson et al., 2002; Bishop & Anderson, 1990; Nehm & Reilly, 2007; Nieswandt & Bellomo, 2009). Although there is some variation in the number of relevant key concepts (e.g. Anderson et al., 2002; Gould, 2002; Nehm & Schonfeld, 2008), at least three concepts – also referred to as core concepts (Nehm & Ha, 2011; Opfer et al., 2012) or principles (Tibell & Harms, 2017) – are both necessary and sufficient to explain evolutionary change by means of natural selection: variation, inheritance, and selection (i.e. differential survival and reproductive success; Endler, 1986; Godfrey-Smith, 2007; Mayr, 2001). Experts’ explanations of evolutionary change are based on core (key) concepts (mentioned above), which are causally central information or features (Opfer et al., 2012). Thus, these core concepts are described as ‘scientifically normative ideas’ (Opfer et al., 2012, p. 753), whereas additional or other key concepts such as limited resources, change over time, and/or speciation (Anderson et al., 2002; Nehm & Schonfeld, 2008; Nieswandt & Bellomo, 2009) are summarised as ‘non-causal features of the evolutionary process’ (Opfer et al., 2012, p. 751). Tibell and Harms (2017) argue that all known key concepts can also be amalgamated under the above-mentioned core concepts. Moreover, concepts such as randomness, probability, temporal and spatial scales (sometimes referred to as threshold concepts) are interlinked with the core concepts and, thus, are particularly crucial for a deeper understanding of evolution (Fiedler et al., 2017; Nehm, 2019; Tibell & Harms, 2017). However, for this literature review, we decided to focus on the three core concepts (i.e. variation, inheritance, and selection) as scientifically normative causal ideas to explain evolutionary change rather than searching for all of the key concepts known from the literature. The following sections describe the three core concepts in more detail.

The core concept variation describes the fact that individuals are genetically diverse, which is manifested as morphological, psychological, and behavioural (phenotypic) differences and similarities. Variation (genotypic or phenotypic) is a pre-requisite for evolution. Without variation, all organisms would have the same traits and natural selection would not occur. The ultimate source of variation is random mutation, genetic recombination, and horizontal gene transfer (Endler, 1986; Mayr, 2001). Although random mutation is often responsible for novel variation within populations, not all mutations give rise to different phenotypes. Mutations may be silent when they ‘occur in non-coding regions of the genome or do not cause any functional changes in gene productions’ (Tibell & Harms, 2017, p. 956). This non-visible genetic variation is challenging to understand compared with obvious phenotypic variation. In fact, learners often have problems incorporating genetic aspects such as random mutations in their explanations of evolution through natural selection (e.g. Bishop & Anderson, 1990; Fiedler et al., 2017; Garvin-Doxas & Klymkowsky, 2008). They also struggle with the importance and nature of random mutations (Garvin-Doxas & Klymkowsky, 2008; Robson & Burns, 2011).

However, for the process of natural selection, (genetic) variation is only relevant if the respective traits are inherited from one generation to the next (the core concept of inheritance or heredity). For example, only mutations that are fixed in the germ cells of sexually reproducing organisms can be passed on to the next generation. Mutations that occur in somatic cells may affect an individual’s fitness (for better or worse), but usually do not alter the fitness of its offspring.¹ Learners seem to misunderstand this inheritance principle and reason that beneficial traits acquired during an individual’s lifetime (whether these are morphological or behavioural changes) are always inherited by offspring (Ferrari & Chi, 1998; Kampourakis & Zogza, 2007).

The third core concept to explain evolutionary change through natural selection is the selection principle itself (also stated as differential survival and reproductive success; Endler, 1986; Mayr, 2001). The physical world imposes numerous biotic and abiotic factors – also called selection pressures (or limited resources) – that limit individual organisms’ survival and determine which phenotypic (and genotypic) traits are beneficial or harmful. In any real environment, these selection pressures are constantly present and change within specific ranges, resulting in different traits conferring different survival potentials and reproductive success. Thus, in a given environment, individuals in a population with certain advantageous traits are more likely to survive and reproduce (i.e. pass on their genes, including the genes for the traits in question) than individuals without those traits. This selection process may involve many generations, and, over long time periods (i.e. deep time), populations may diverge to become separate species that can no longer interbreed (also known as the origin of species or speciation; Tibell & Harms, 2017). The selection principle is often referred to as survival of the fittest, which is misleading for several reasons (Gregory, 2009). For example, learners often misinterpret this biological phrase with their everyday conception of a fitness condition (i.e. being physically fit and healthy). Thus, learners think that only the most suitable individuals survive (Gregory, 2009).

Core concepts in evolution within science standards

Science standard frameworks have introduced the notion of disciplinary core ideas that are woven across learning contexts to support a continuous integration of knowledge and ability over multiple years (National Research Council [NRC], 2012; Sekretariat der Ständigen Konferenz der Kultusminister der Länder in der Bundesrepublik Deutschland [KMK], 2005). In the life sciences, evolution is the most important organising theme, and national standards in several countries consider aspects of evolution as core ideas (e.g. Department for Education [DfE], 2015, NRC, 2012, KMK, 2004, 2005). For instance, the Next Generation Science Standards (NGSS Lead States, 2013) lists ‘biological evolution: unity and diversity’ (p. 1) as a core idea of life sciences, while other aspects of evolution are implemented in the core idea ‘heredity: inheritance and variation of traits’ (p. 1). Teaching evolution should be implemented in secondary school education as a coherent logic of adaptation by natural selection (e.g. DfE, 2015; NRC, 2012; NGSS Lead States, 2013; KMK, 2004). This curricular logic is underpinned by research from cognitive science showing that knowledge integration and recall is promoted if the information is structured along with core concepts of the field (e.g. Clark & Linn, 2003; Opfer et al., 2012). Biologists’ concepts of plants and animals facilitate classification and, therefore, are core concepts in taxonomy (Opfer et al., 2012). However, up to the age of 8 years, children’s understanding of such core concepts is seldom scientifically correct and often differs from adults’ concepts (for an overview, see Inagaki & Hatano, 2004; Wellman & Gelman, 1992). For example, children are less likely to attribute features of life to plants than to animals (concept of living/non-living; Coley et al., 2002; Richards & Siegler, 1986) and they consider plants in contrast to animals as non-active beings (Carey, 1988). Between the ages of 7 and 9, children start to understand plants as living things (Hatano & Inagaki, 1994). However, children still refer to human bodily processes when explaining biological phenomena in plants and animals, hence using vitalistic concepts as a point of reference (Inagaki & Hatano, 2004). Overall, promoting children’s understanding of scientifically correct core concepts is one of the major aims of science education (NRC, 2012), because the curricular logic of core concepts should support knowledge integration and recall.

Cognitive biases in evolution education

Difficulties in learning evolution stem from cognitive biases that coexist with central and scientifically correct concepts and persist over time (Opfer et al., 2012). Resistance to scientific explanations originates in early childhood, as children develop intuitive assumptions about the biological world known as ‘naïve biology’ (see also ‘naïve physics’ and ‘naïve psychology’; Wellman & Gelman, 1992, p. 337), that is, ‘what children know before their exposure to science’ (Bloom & Weisberg, 2007, p. 996). Children’s reasoning about the biological world and, later on, students’ explanations of evolution (Opfer et al., 2012) are mainly influenced by three cognitive biases: essentialism (e.g. Shtulman & Schulz, 2008), teleology/anthropomorphism (e.g. Kelemen, 1999), and intentionalism (e.g. Inagaki & Hatano, 2004; see Table 1).

Table 1. An overview of the three cognitive biases (i.e. essentialism, teleology, and intentionalism) and how they impede understanding of evolutionary concepts

Cognitive bias	Impedes understanding of:
Essentialism	variation, because within-species variation is overlooked (Shtulman & Schulz, 2008) species as individuals within populations that are affected differently by the environment, because species are perceived as biological kinds (Shtulman, 2006)
Teleology	randomness (variation), because novel traits arise in response to need or selective pressure (e.g. Kampourakis & Zogza, 2007) variation, when a trait is applied equally to all individuals of a population because the trait fulfils a purpose (e.g. Emmons & Kelemen, 2015)
Intentionalism	populations evolving instead of individuals reacting, because an individual’s intention to adapt is emphasised (Moore et al., 2002) probabilistic processes, because nature is perceived as a conscious agent leading to evolutionary change (e.g. Gregory, 2009; Kampourakis & Zogza, 2007)

The essentialist bias (or psychological essentialism; Kampourakis, 2015) is characterised by the belief that category members (e.g. animal species) share an intrinsic nature or essence (Shtulman & Schulz, 2008) that is passed down (inherited) to the offspring and will not change as a result of the environment or upbringing (Gelman & Wellman, 1991). This essence is defined by the shared characteristics of category members (sortal essence) or the causes that determine category membership (causal essence; Gelman, 2003). Evidence shows that children’s essentialism stems from three possible sources: (1) children reason about the properties of an animal as yet unknown to them based on the animal’s identity (i.e. category membership) and not perceptual similarity (Gelman & Coley, 1990); (2) an animal’s category membership stays constant during superficial changes (Rosengren et al., 1991); and (3) an animal’s category membership stays constant during changes in the environment of upbringing such as parent animals of a different kind (Waxman et al., 2007; Gelman & Wellman, 1991). Although essentialism is sufficient for children’s reasoning about the biological world in some respects (e.g. stability of traits in the case of adoption), it impedes understanding of evolution for several reasons (Gelman & Rhodes, 2012, p. 14): (1) stability and immutability of species hinders their change; (2) ‘sharp boundaries’ between categories prevent ‘intermediate categories’; (3) variation only concerns ‘superficial features’; (4) causes are inherent to individuals and so are changes; and (5) evolution progresses towards species’ ‘ideal’ forms.

In contrast, the teleological (and anthropomorphic) bias stems from a design stance in nature (design teleology; Kampourakis, 2015, 2020), assuming a purpose, need, or even a plan as a mechanism of change (Lennox & Kampourakis, 2013). Teleologically biased explanations consider the contribution of phenomena in organisms and natural objects (i.e. their function) to a final end (i.e. their purpose). Hence, teleological stances in evolution understanding consider the development of traits as purposeful and oriented towards a final goal (Nehm & Reilly, 2007; Nehm & Schonfeld, 2007; Kampourakis & Zogza, 2008). Anthropomorphically biased explanations reflect the needs of individuals and, hence, trait changes follow the agency of individuals to cope with environmental changes (Gregory, 2009; Moore et al., 2002). For a long time, children were regarded as ‘promiscuous’ teleologists in that they ascribe a purpose to the genesis of natural objects and organisms as well as artefacts (Kelemen, 1999, p. 1441). However, in the transition from pre-school to school, teleological explanations are nuanced in favour of organisms and artefacts but not for natural objects (Kampourakis et al., 2012). Students prefer teleological explanations (Trommler et al., 2017), and even teachers use teleological accounts when introducing randomness into the classroom as an explanation for evolutionary mechanisms (Gresch & Martens, 2019).

The third cognitive bias, the intentionalist bias, considers the mental state of an individual as an explanation for the cause of a phenomenon. With an intentionalist bias, an individual’s intention to adapt is emphasised (Moore et al., 2002), that is, individuals react to environmental changes rather than selective pressure(s) acting upon whole populations. On the other hand, nature is also perceived as an intentional agent of evolutionary change, hence, impeding the understanding of randomness and probability in evolutionary change (Gregory, 2009; Kampourakis & Zogza, 2007). The intentionalist bias stems from children’s naïve biology, which facilitates children’s predictions of natural phenomena through different causal devices – in the case of intentionalism through an individual’s intentional action (Inagaki & Hatano, 2004). Furthermore, Gregory (2009) distinguishes between internal intentionalism (i.e. related to individuals’ intentions) and an external intentionalism referring to a conscious agent (e.g. nature; Kampourakis & Zogza, 2007). Although teleology and intentionalism are closely related (i.e. functions are intentionally designed; Kelemen, 1999), they are distinguishable (Evans, 2001). Teleological bias results from a design stance but is not necessarily intentional.

Cognitive biases constrain understanding of evolution

Cognitive biases are usually considered to be constraints to learning, although when made explicit they can promote it (Evans et al., 2012; Opfer et al., 2012). Cognitive biases can constrain learning at three different levels: the organism (i.e. the level of the learner), the task, and the environment (i.e. the surroundings or culture the learner lives in; Rosengren et al., 2003). First, cognitive biases are rooted in the individual because they develop from intuitive assumptions, and, therefore, are related to the organism level. Second, cognitive biases often interact with constraints at the task or environmental level (Rosengren et al., 2003). For example, cognitive biases such as essentialist, teleological, or intentionalist reasoning are applied depending on the context of macro- or microevolution (i.e. task level; Rosengren & Evans, 2012). Thus, constraints at the three levels often interact with each other. In addition, the function of cognitive biases differs depending on the context of the task (e.g. animal vs. plant, trait gain vs. trait loss, between- vs. within-species changes; Nehm & Ha, 2011; Opfer et al., 2012). For instance, undergraduate students use more key concepts to explain evolutionary changes concerning animals than plants, while the amount of cognitive biases does not differ (Opfer et al., 2012). Cognitive biases seem to be ingrained over different evolutionary contexts, yet are independent from using scientifically accurate concepts.

Coley and Tanner (2015) demonstrated how undergraduate students’ misconceptions of evolution arise from what they called cognitive construal, such as teleological, essentialist, and ‘anthropocentric’ [sic!] thinking (p. 11). Hence, confronting cognitive biases in the early stages of education is potentially a way of accounting for different levels of constraints such as organism or task contexts (Evans et al., 2012; Kelemen, 2012; Shtulman & Calabi, 2012). The interactions of organism, task, and environment constraints result in different learning progression pathways (Rosengren & Evans, 2012). Thus, the analysis of organismic constraints in early childhood development provides a learning progression that accounts for cognitive biases. Here, children’s intuitive assumptions on evolutionary mechanisms provide stepping stones (Evans et al., 2012), intermediate levels that bridge children’s learning progression from lower building blocks (i.e. precursory concepts) towards higher building blocks (i.e. scientifically correct concepts; Duschl et al., 2011; Scott et al., 2019).

Research questions

Despite the extent of findings on misconceptions and biases throughout formal education, as well as literature reviews documenting the current state of research on evolution education in schools (e.g. Glaze & Goldston, 2015; Gregory, 2009; Russell & McGuigan, 2015; Ziadie & Andrews, 2018), an empirical review of precursory concepts of evolution before the age of 7 is missing. Based on the theory of children’s development of precursory concepts regarding evolution (e.g. Evans et al., 2012) and cognitive biases (e.g. Wellman & Gelman, 1992) as described above, the following overarching research question guided our analysis of articles in this systematic literature review.

Which precursory concepts in evolution do children in early childhood and primary education (aged up to 7 years) (a) already possess or (b) develop after educational intervention?

Our research question was refined by using the following criteria regarding the population, outcomes, study design, and type of intervention (i.e. structured ‘question’; Khan et al., 2003, p. 119) to guide our analysis.

• Population: children aged up to 7 years (includes early childhood and early primary education)

• Outcomes: precursory concepts and knowledge about evolution

• Study design: (quasi-)experimental and interview studies

• Interventions: formal (i.e. pre-school, primary/elementary school, kindergarten) and informal (i.e. museum, home) learning opportunities

Methodology

We conducted a systematic literature review by searching three databases representing different disciplines (i.e. social science, education, and psychology) and analysing the resulting data set. As reviews of literature are limited by the subjectivity of the researcher, we documented the research process in detail, including a definition of the search terms, the selection of papers for analysis, and the analysis itself (see Figure 1). This ensures transparency and replicability, as illustrated in other reviews (e.g. Glaze & Goldston, 2015). We used the Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA) checklist, which provides guidelines for reporting systematic reviews in order to comply with standards and enhance replicability (Moher et al., 2009). All the papers included in the analysis were read several times and reviewed using Critical Review Forms (Law et al., 1998a, 1998b), as commonly used for systematic literature reviews (e.g. Glaze & Goldston, 2015).

Figure 1. Flow-chart of the search procedure and data analysis used for the systematic literature review, following PRISMA (Moher et al., 2009)

Search procedure

We selected databases that represent different disciplines that address evolution education, science and education, and psychology: Web of Science by Thomson Reuters (social and natural sciences, humanities), Education Resource Center (ERIC) by the US Department of Education (educational science), and PsychINFO by the American Psychological Association (psychology). The databases were searched on 29 May 2019. To obtain relevant articles, the following criteria were used to formalise our search: (1) relevant search terms were used; (2) the publications were peer reviewed; (3) the publications were articles; and (4) the articles were written in English. Initially, we did not limit the publication date (the earliest paper was published in 1977). However, to avoid a bias towards research on children’s understanding of inheritance that occurred during the search, we subsequently limited our in-depth analysis to articles published after 2007.

The search terms (keywords) and criteria were defined to generate an initial database that included all potential articles relevant to the objectives of the literature review. However, while the keywords should represent as many concepts related to the focus objective as possible (Rönnebeck et al., 2015), the amount of literature to be reviewed needs to be manageable. Our keywords were intended to represent the different concepts used in evolution education, such as macroevolution and microevolution. The keyword ‘evolution’ may be too general to be used as a single keyword for biological evolution, because it can also relate to more general developmental processes (e.g. evolution of degree and direction of hand preference in children), so to specify the search in terms of evolutionary theory, we added the descriptor ‘biol*’ (the asterisk indicating word truncation). Microevolutionary concepts such as variation, inheritance, and (natural) selection also served as search terms. In addition, keywords were used to specify the relevant age group (i.e. pre-schoolers). Altogether, the following search terms were used in various combinations: variation, inheritance, natural selection, biol* evolution, kindergarten, early childhood, pre-school (education), elementary, and primary. The following are examples of the full search expressions used in the ERIC database:

• evolution AND (biology OR biological) AND (educationlevel: “early childhood education” OR educationlevel: “kindergarten” OR educationlevel: “preschool education”)

• “natural selection” AND (educationlevel: “early childhood education” OR educationlevel: “kindergarten” OR educationlevel: “preschool education”)

• variation AND (biology OR biological) AND (educationlevel: “early childhood education” OR educationlevel: “kindergarten” OR educationlevel: “preschool education”)

• inheritance AND (biology OR biological) AND (educationlevel: “early childhood education” OR educationlevel: “kindergarten” OR educationlevel: “preschool education”).

Data analysis

The empirical findings retrieved on learning evolution before the latest possible transition from early childhood education (ISCED 0) to primary education (ISCED 1; i.e. age 7, including comparison groups that were older than 7 years) were then analysed. The analysis was guided by the results section of the checklist used for reporting systematic reviews: it included the reporting of study characteristics, risk of bias within studies, and study results (Moher et al., 2009). We used the Critical Review Form for Qualitative Studies (Law et al., 1998a) and the Critical Review Form for Quantitative Studies (Law et al., 1998b) for consistent analysis between the articles and to document the analysis procedure for each article. The Critical Review Forms provide comparable criteria for quantitative and qualitative studies to extract the main characteristics of each study regarding study rationale and purpose, sample description, observations and effect sizes, measures to achieve reliable and valid interpretations, and connections to previous literature. Articles were first screened based on their title and abstract to exclude articles not related to the study objectives. Then, full-text articles were assessed to select eligible articles based on their topical focus and age group. The included articles were then read several times, reviewed according to the criteria of the Critical Review Forms, and finally documented (documentation is available upon request). Discussion between the researchers based on duplicate reviews of several articles ensured the consistent application of the Critical Review Forms.

Results

The database searches resulted in N = 204 individual articles (after n = 56 duplicates had been eliminated). These 204 articles were then screened based on their title, abstract, and content. We subsequently excluded articles that were (1) not related to the topic of evolution education (n = 121; e.g. neuropathology, child development), (2) related to evolution education but for different age groups (n = 33; e.g. secondary education, science teacher education), or (3) related to early education but not to the above-mentioned core concepts in evolution (n = 6; e.g. fossils). Eighteen further articles had to be excluded because they did not present empirical data (i.e. they were comments on previous articles, textbook analyses, theoretical papers, or activity descriptions). As a result, n = 26 articles were obtained for further analysis (variation, n = 3; inheritance, n = 17; natural selection, n = 6). We then limited our in-depth analysis to articles published after 2007, to avoid the strong bias towards the inheritance concept (n = 17), because research on children’s precursory concepts before 2007 mainly focused on the core concept inheritance. Our final data set therefore comprised n = 3 articles on variation, n = 8 articles on inheritance, and n = 6 articles on natural selection.

Children’s understanding of variation

All three analysed articles focusing on the core concept variation (Emmons & Kelemen, 2015; Herrmann et al., 2013; Shtulman & Schulz, 2008) investigated to what extent children (5−8 years) endorse within-species variation as a foundation of adaptation by natural selection (Table 2). One study examined children’s endorsement of individual change within a life-span (i.e. whether children understood biological mechanisms of life-span change; Herrmann et al., 2013), while the other two studies explored how far children endorse variation between individuals of one species (i.e. within-species variation; Emmons & Kelemen, 2015; Shtulman & Schulz, 2008).

Table 2. An overview of the empirical studies focusing on the core concept variation in early childhood evolution education

Study	Sample Age group; sample size; mean (M) age (SD/range)	Intervention/procedure	Design & measurement	Results	Conclusions
Shtulman and Schulz (2008)	4−6 yrs; n = 22; M = 6;6 yrs (4;2–6;7) 7−9 yrs; n = 21; M = 8;3 yrs (6;9–9;11) Adults (divided by post-test): transformationist, n = 19; M = n/a variationist, n = 15; M = n/a	Interview along with a picture book on six animals with three traits each and judgement task for each: trait presentation (generic question) and function description; actual variability question; if trait common to all members, then potential variability question	Design: experimental within-between-subject design (factors: animal type and trait type as well as participant group) Measurement: judgement task (actual and potential) and justification questions: interview for children, supported by picture book showing a single individual of each animal; questionnaire for adults: adults were also divided into two groups (transformationists, variationists) by post-test (evolution comprehension assessment; Shtulman, 2006)	Judgement patterns: adult judgements correlated with comprehension (actual: r = .53; potential: r = .64) Variationist adults more often endorsed actual variability Main effect of participant group and trait type: variationist adult > transformationist adult > child variationist adult distinguished internal vs. non-internal others distinguished behaviour vs. non-behaviour	Psychological essentialism prevents appreciation of within-species variation Essentialism is negatively correlated with knowledge of natural selection With age, appreciation of actual variability increases
Herrmann et al. (2013) – Study 1	3 yrs; n = 26; M = 47.2 mos (39−51) 4 yrs; n = 24; M = 49.9 mos (46−53) 7 yrs; n = 26; M = around 7 yrs (range n/a)	Eight trials (different animals) on four change types (growth, + colour/texture, + shape, species): (1) introduction on actual status of juvenile form (2) affirmative question on two adult forms (subsequently presented) (3) if affirmed, then two biological and three non-biological reasoning questions	Design: cross-sectional experiment (within-subject: change and reasoning type; between-subject: age) Measurement: 16 affirmative questions; 16 yes-or-no follow-up questions on biological reasoning; 24 yes-or-no follow-up questions on non-biological reasoning	Main effects: reasoning type, more biological reasoning (η^2 = 0.91), change type, less acceptance of species change (η^2 = 0.64), age (η^2 = 0.12) age × reasoning type interaction effect: less non-biological reasoning, but more biological reasoning for 7-year-olds (η^2 = 0.41) age × reasoning × change interaction effect: younger children gradually endorse change, no difference for older children (η^2 = 0.13)	Endorsement of change is age dependent as 3- and 4-year-olds gradually endorse change depending on change type and older children (7 yrs) equally endorse change More biological reasoning on changes for all age groups, but dependent on change types 3- and 4-year-olds used less biological reasoning for dramatic changes (grow + texture/colour + shape)
Herrmann et al. (2013) – Study 2	3 yrs; n = 14; M = 46.4 mos (40–49) 4 yrs; n = 16; M = 53.0 mos (51–56) control: n = 14; M = 52.1 mos (47–58)	(1) Data from study 1 as pre-test (2) Intervention on 3- and 4-year-olds: two terrariums (pregnant mouse or caterpillar) observed over 2 mos (3) Post-test control group got no intervention	Design: pre-post intervention with control group Measurement: same as study 1	Main effect time: more reasoning at post-test (η^2 = 0.20) Main effect reasoning: more biological than non- biological reasoning (η^2 = 0.88) Main effect type of change: more possible than impossible (η^2 = 0.85) Interaction effect reasoning × type of change × test time (η^2 = 0.10): increase for biological reasoning on possible changes, stable for non-biological reasoning and impossible changes No change in control group	Witnessing the life-span change promotes acknowledgement of the biological nature of the process and therefore changes reasoning on this category of change
Emmons and Kelemen (2015) – Study 1	5−6 yrs; n = 20; M = 5;11 yrs (SD = 4 mos) 7−8 yrs; n = 20; M = 8;1 yrs (SD = 8 mos)	Stimuli on four fictional animals (two mammals, one bird, one fish) with three trait types (internal, external, and behavioural): (1) introduction to novel animal (unfamiliar and singular) (2) trait and trait function provided (3) forced-choice question	Design: quasi-experimental study (between-subject factor: age; within-subject: trait type; counterbalanced design: two animals per child Measurement: 12 forced-choice questions (4 animals × 3 trait types) with two answer options: essentialist or variationist	No main or interaction effects of age or trait type on variation endorsement	For 5−8-year-olds only marginal tendency of choosing trait homogeneity over within-species variation Higher variation acceptance than in prior research (e.g. Shtulman & Schulz, 2008) Variation acceptance is rather context dependent (not biases like essentialism)
Emmons and Kelemen (2015) – Study 2	5−6 yrs; n = 20; M = 5;11 yrs (SD = 7 mos) 7−8 yrs; n = 20; M = 8;0 yrs (SD = 6 mos)	Same as study 1, but in step (2) trait function omitted	Same as study 1	Main effect age: higher variation endorsement in older compared with younger children (p = .03, ${\eta }_p^2$ = 0.12)	Older children (7–8 yrs) accepted variation more than younger children (5–6 yrs)

Note. yrs = years; mos = months; n/a = not available; n.s. = not significant; OR = Odd’s ratio

The aim of the study by Herrmann et al. (2013) was to investigate age and the impact of knowledge growth on essentialist bias and the endorsement of change in individuals. The authors predicted that increasing knowledge about change in individuals (i.e. metamorphosis) would positively influence the appreciation that an individual’s category identity stays stable even if surface features dramatically change. This is because of the innate potential of organisms to change and the immutability of categories, two main components of essentialist bias (Gelman, 2003). Therefore, children from three age groups (3, 4, and 7 years old) were introduced to the juvenile forms of different animal species and had to affirm two adult forms. The required changes were possible in three cases (growth, growth and colour/texture, growth and colour/texture and shape), while a fourth case required a species change (e.g. dog to cat). If children regarded the change as possible, they answered two biological and three non-biological reasoning questions. The results revealed that all children reasoned biologically about life-span changes and did not accept the species change. However, older children (7 years old) endorsed life-span changes equally, whereas younger children endorsed greater changes less. To investigate whether the age differences were the result of an increase in knowledge, the authors conducted an intervention that promoted the observation of life-span changes (i.e. changes of growth and changes of growth, colour/texture, and shape) and its influence on the biological reasoning of life-span changes. The results showed that children were more likely to reason biologically regarding the three growth change types, while reasoning for the impossible species change decreased. The authors concluded that first-hand observation of changes promotes knowledge about possible changes over a life-span. Therefore, the authors suggested a developmental progression of essentialist reasoning: with older age and more knowledge, children’s stronger belief in innate potential and immutability allows them to accept more dramatic changes in individuals. But the accepted changes are still restricted within the species.

The constraints of the essentialist bias to within-species changes prevent children from accepting more dramatic changes. Immutability is one reason why creationism is preferred over evolutionary concepts (Evans, 2000; Samarapungavan & Wiers, 1997). However, Shtulman and Schulz (2008) tested whether an essentialist bias is related to knowledge about evolution and whether adults endorse within-species variability to a stronger degree than children do. Children (4–9 years old) and adults were presented with six animals and three traits (i.e. behavioural, external, and internal), and the stimuli included trait function information. For each animal and trait (6 × 3), the experimenter asked the participants whether they regarded this trait as actually variable among species members and, if not, whether it was potentially variable. For traits seen as actually and potentially not variable, participants were asked for their reasoning. Evolution assessment scores were higher for those adults who judged the traits to be variable (actually: r = .53; potentially: r = .64). Comparing adults and children, participant group and trait type were the main factors influencing the endorsement of actual variability: adults with a variationist view of evolution most often endorsed variability, while children endorsed variability the least. For trait type, behavioural traits were more often judged as variable than non-behavioural (internal and external) traits. A similar pattern of judgement held true for potential variability, but the children showed a strong essentialist bias.

Building on the two previous studies, Emmons and Kelemen (2015) investigated the influence of children’s age and trait type on their endorsement of the concept of variation. The authors criticised the study of Shtulman and Schulz (2008) for questioning the reasoning if specific traits were potentially variable. Emmons and Kelemen (2015) argue that ‘the question may have led children to misconstrue it (i.e. they do not understand the question as asking whether species members can be born with a different trait than their parents)’ (p. 150). Furthermore, the authors experimentally varied trait function information. They investigated whether information on function and purpose (as per educational recommendations on teaching trait functions; NGSS Lead States, 2013) prevents children from endorsing within-species variation. For this, children from two age groups were shown four animals (two per child in a counterbalanced design). Each animal possessed one of the three functional trait types (i.e. behavioural, external, and internal). After each animal and its trait type had been explained, the children were asked whether all animals of that kind share the trait. When the trait function was included, children’s endorsement of variation occurred no more frequently than by guessing (irrespective of age group). When the trait function was omitted, older children (7–8 years old) endorsed variation more than by chance, while younger children (5–6 years old) remained essentialist or uncommitted reasoners.

In summary, the review of the three studies focusing on variation indicated that children’s abilities to recognise within-species variation depend on their age and contextual factors such as trait type and trait function information. Thus, children of about 7 years of age not only know about possible life-span changes (growth, colour/texture, and shape) but can also identify within-species variation. Children’s beliefs in innate potential increases with knowledge about life-span changes (Herrmann et al., 2013). Around 7 years of age, trait function information promotes teleological beliefs resulting in essentialist reasoning that functional traits are stable (Emmons & Kelemen, 2015). The lack of trait function information positively influences the acceptance of trait variability, but only for older children (7–8 years old; Emmons & Kelemen, 2015). In addition, trait types such as behavioural and physical traits moderate the endorsement of variation, with children believing that behavioural traits are more variable (Shtulman & Schulz, 2008).

Children’s understanding of inheritance

Most of the articles in our data set investigated children’s understanding of the core concept inheritance (n = 17), as there is a long history of research on children’s understanding of inheritance (e.g. Wood-Robinson, 1994, for a review). Some of the articles provided an overview of the evidence from 1989 to 2003. To avoid a bias of articles on inheritance, we limited the publication dates to be comparable with the two other categories (i.e. only articles between 2007 and 2019 were analysed in depth), resulting in n = 8 articles (Table 3).

Table 3. An overview of the empirical studies focusing on the core concept inheritance in early childhood evolution education

Study	Sample Age group; sample size; mean (M) age (SD/range)	Intervention/procedure	Design & measurement	Results	Conclusions
Raman (2018) – Study 1	4–5.5 yrs; n = 29; M = 4;8 yrs (4;1–5;2) adults; n = 47; M = 21;5 yrs (18;5–25;5)	Eight vignettes (modified ‘switched-at-birth’ tasks) that focus on height (4 × tall birth parents, short adoptive parents; 4 × short birth parents, tall adoptive parents) or on weight (4 × thin birth parents, fat adoptive parents; 4 × fat birth parents, thin adoptive parents)	Design: interview-study, 2 (age) × 2 (type) ANOVA Measurement: scoring 0 for incorrect and 1 for correct parent matching	Main effect type: more correct matches for height than weight (p < .01) Main effect age: adults made more correct matches (p < .01) Interaction effect age × type: both groups made more correct matches for height than weight (p < .05)	There is a developmental difference, with adults attributing more correct matches to height than children do
Raman (2018) – Study 2	4–5.5 yrs; n = 24; M = 4;6 yrs (4;0–5;1) 6–7 yrs; n = 25; M = 6;8 yrs (6;0–7;3) 8.5–10 yrs; n = 25; M = 9;3 yrs (8;8–9;7) 10–11 yrs; n = 24; M = 11;3 yrs (10;9–11;10) adults; n = 24; M = 22;6 yrs (18;3–26;10)	Six vignettes (modified ‘switched-at-birth’ tasks) that focus on the influence of food on weight (3 × healthy food for baby, birth parents fat, adoptive parents thin; 3 × unhealthy food for baby, birth parents thin, adoptive parents fat)	Design: interview-study, 5 (age) × 2 (food) ANOVA Measurement: scoring 0 for incorrect and 1 for correct parent matching, with either option scored as 0.5. In a second conservative analysis the either option was scored as 0	Main effect age: 4–5.5-year-olds made more correct matches than 6–7-year-olds, 10–11-year-olds, and adults (ps < .01); 8.5–10-year-olds made more correct matches than all others (ps < .01) Interaction effect food × age: for healthy food, 8.5–10-year-olds made more correct matches than all others (p < .01) Conservative analysis: interaction effect food × age: for healthy and unhealthy food, 4–5.5-year-olds and 8.5–10-year-olds made more correct matches than all others (ps < .01)	8.5–10-year-olds were likely to identify inheritance as determining weight, more than any other group There was a sharp decline in the perceived role of inheritance in the determination of weight for 10–11-year-olds and adults
Raman (2018) – Study 2a	4–6 yrs; n = 19; M = 4;9 yrs (4;1–5;6) adults; n = 23; M = 23;2 yrs (18;6–28;3)	Six vignettes (modified ‘switched-at-birth’ tasks) that focus on influence of food on height (3 × healthy food for baby, birth parents short, adoptive parents tall; 3 × unhealthy food for baby, birth parents tall, adoptive parents short)	Design: interview-study, 2 (age) × 2 (food) ANOVA Measurement: scoring 0 for incorrect and 1 for correct parent matching	Main effect age: adults made more correct matches than 4–6-year-olds (p = .01)	Neither age group differentiated between the effects of eating healthy or unhealthy foods for the determination of height Children might have a rudimentary framework to recognise that biological processes such as height are dictated by genetics, but it is highly unlikely that they have the same kind of understanding as adults do of the underlying mechanisms of how genes affect height
Moya et al. (2015)	Peru; n = 193; M = 26 yrs (4–75) Fiji; n = 119; M = 27 yrs (5–73) United States; n = 302; M = 32 yrs (18–64)	Adoption vignettes that were modified for each cultural context Two vignette conditions: adoption vignette – boy born to a set of parents and raised by another set migration vignette – boy is raised by his biological parents who migrate from group A into group B when he is an infant Two group conditions: ingroup – both the biological parents and the cultural models are drawn from the same group intergroup – biological parents and the people who served as cultural models (i.e. adoptive parents or cultural context), are identified as belonging to different social groups	Design: experimental study with two between- (field site; group condition) and one within- (vignette condition) subject factors Measurement: protocol	Vignette condition: adoption: regression model with trait type, socialisation index and interaction (adoption vignette), i.e. younger children fail to differentiate the two kinds of traits and they show a slight birth bias for all traits (choosing a birth parent resemblance around 60–80%) migration: pattern persists Group condition: Only US participants showed a modest belief that vertical cultural transmission occurs	Morphological traits are under pre-natal influence whereas belief traits are culturally influenced However, cultural influence differs between participant groups, i.e. trait transmission is influenced by ‘human’s evolutionary history as cultural species’
Ergazaki et al. (2015)	5−5.5 yrs; n = 60; M = n/a	Three part teaching intervention (1−1.5 h on each of three days): (1) six activities on reproduction and mechanism of species resemblance; (2) two activities on mechanism of resemblance in body traits; (3) two activities in consolidation phase	Design: mixed-model case study with pre–post semi-structured interviews for designing and evaluating a teaching intervention Measurement: interview protocol comprising three parts: reproduction (task 1), parents–offspring family resemblance (task 2.1 and 2.2), parent–offspring body trait resemblance (task 3.1 and 3.2); analysis of transcribed answers by ‘open coding’ scheme (κ = 0.90)	Pre–post improvement in recognising the contribution of both parents for reproduction (ps < .0001) Pre–post improvement in justifications for species resemblance (ps < .0005) Pre–post improvement in consistency of claims for parents’ body trait resemblance (p = .0015) and biological justifications (p < .0001) Pre–post improvement in consistency of claim for wish unfulfilability (p = .0002) and biological reasoning (p = .0019)	Inheritance concept is not too abstract for 5-year-old children if genes are introduced as essence-like species-genes
Ergazaki et al. (2014)	4.5–5.5 yrs; n = 90; M = n/a	–	Design: exploratory case study with semi-structured interviews Measurement: interview protocol comprising five phases with multi-case versions of tasks on adoption, functionality, acquired traits, wish-fulfilment and mechanisms; analysis of transcribed answers by coding for claims, justifications and trait transmission (κ = 0.90)	Tendency for a ‘kinship-bias’: children attributed traits more often to biological parents (vs. step-parents; p < .01), especially behavioural traits (vs. physical; p < .05); Differential justification pattern for trait attribution to biological vs. step parents: birth and biological family resemblance equally for biological parents; nurture less frequently than step-parents resemblance for step-parents Children prefer ‘pre-biological model’ over ‘mechanical model’ for explanations on kinship (p < .05)	Children do not differentiate between trait types (behavioural vs. physical) to be inherited; however, wrong but inheritance-based explanations are more useful than correct but psychologically based claims
Kaminski et al. (2013)	5 yrs; n = 30; M = 4.8 yrs (4.5–5.3) 7 yrs; n = 23. M = 6.8 yrs (6.4–7.3) 9 yrs; n = 27. M = 8.9 yrs (8.4–9.9) 11 yrs; n = 20. M = 10.9 yrs (10.4–11.5) adults; n = 29; M = 20.8 yrs (18–25)	12 items per condition (control vs. experimental), each with one picture of a newborn face on top (12 boys/12 girls) plus three pictures of female faces under it of which one shows the mother of the newborn: newborn’s face + genetically related face Control condition: + two unrelated faces with non-kin (sharing no perceptual similarity with newborn face) Experimental condition: + non-kin face perceptually related to the newborn’s face (sharing a salient perceptual feature with the target which was irrelevant to kinship) + unrelated face with non-kin face (sharing no perceptual similarity with newborn face)	Design: experimental study: between-subject design (control vs. experimental: features; age: 5, 7, 9, 11, adult) Measurement: participant’s choice for each item (BMDR and distracter score); scoring 0 for incorrect and 1 for correct mother–newborn matching	In control condition, BMDR higher than chance (36.5%, p = .009) but affected by age (p < .0001): different from chance by the age of 9 yrs (ps < 0.05); performance increase between ages of 7 and 9 yrs (OR: 1.43, p = .042) In experimental condition, BMDR not different from chance (34.5%) but affected by age (p < .0001) and age × feature (p = .024): different from chance by the age of 11 yrs (ps < .05); performance increase between the ages of 7 and 9 yrs (OR: 1.47, p = .056)	Difficulties in detecting parent–child resemblance stem from lack of knowledge about inheritance for 5- and 7-year-olds An additional factor, however, explains changes between 7 and 9 yrs: knowledge-based processes are accompanied by perceptual difficulties in differentiating relevant and irrelevant properties
Williams (2012)	4−5 yrs; n = 46; M = 5.5 yrs (3;10 −5;2) 7−8 yrs; n = 45; M = 6.9 yrs (6;1−7;9) 10−11 yrs; n = 45; M = 10.8 yrs (9;7−11;6) 14−15 yrs; n = 46; M = 14.1 yrs (13;8−14;8)	Six vignettes for each task (phenotypic similarity task and phenotypic difference task) with drawings depicting families with three feature types (physical and personal features as well as disabilities) and two items per feature	Design: experimental within-(features: 3; offspring: 2) between- (age: 4) subject design; individual interviews in school setting Measurement: phenotypic similarity task with judgements (fixed choice questions; 100% IRR) and explanations (open-ended questions; 90% IRR); phenotypic difference task with judgements (100% IRR) and explanations (90% IRR)	4−5-year-olds: no consistent mother bias, but no explanations for offspring traits; sibling possess same traits as offspring, but no explanation for resemblance; no explanation for phenotypic dissimilarity; no explanation for traits of second-generation offspring 7−8-year-olds: enhanced mother bias and explanations on offspring resemblance did not involve both biological parents; judged that siblings possess different traits, but gave no explanations; ‘non-inheritance’ explanations for phenotypic difference and second-generation resemblance 10−11 and 14−15-year-olds: few developmental changes between 10 and 14 years old; preferred influence of both parents on offspring’s traits and used genetic explanations; sibling might inherit similar traits as the offspring, explanations were mostly based on immediate kinship and genetics; phenotypic difference was explained by genetic reasons, as second-generation traits were	4-year-olds are able to make basic inheritance judgements (implicit knowledge) but the ability to verbalise explanations explicitly is delayed to older than 7 yrs (explicit knowledge)
Williams and Smith (2010)	4−5 yrs; n = 46; M = 5.5 yrs (3;10−5;2) 7−8 yrs; n = 45; M = 6.9 yrs (6;1−7;9) 10−11 yrs; n = 45; M = 10.8 yrs (9;7−11;6) 14−15 yrs; n = 46; M = 14.1 yrs (13;8−14;8)	Vignettes focused on human adoption	Design: experimental within-(features: 3) between- (age: 4) subject design; individual interviews in school setting Measurement: adoption task on kindhood judgements (forced-choice questions) and explanations (100% and 92% IRR); causal mechanism task (fixed-choice methodology); family relatedness task on judgements and explanations (90% IRR)	Main effect age (judgement task): increasing age was linked to more differentiated response patterns (depending on feature type; p < .001) Main effect age (explanations): 4-year-olds gave any explanation, 7-year-olds explained by immediate kinship, 10- and 14-year-olds gave similar higher level inheritance explanations Main effect age (causal mechanism): 4- and 7-year-olds endorsed the ‘developmental’ mechanism more than either the 10- or 14-year-olds whereas 10- and 14-year-olds endorsed ‘inheritance correct’ mechanism more than 4- and 7-year-olds (ps < .05) Main effect age (family relatedness judgements): more selective judgements associated with higher age (p < .001; 84.8% of 4-year-olds were non-selective in judgements of relatedness)	4-year-olds have implicit knowledge of inheritance (inheritance of physical features, internal biological mechanism) 7-year-olds are introduced to biological concepts of inheritance mechanisms 10- and 14-year-olds conceptualise family relatedness but have no knowledge of genetic transmission
Waxman et al. (2007) – Experiment 1	4–5 yrs; n = 33; M = n/a 6–7 yrs; n = 47; M = n/a 9–10 yrs; n = 55; M = n/a Adults; n = 34; M = n/a	Three adoption scenarios (cow–pig, pigeon–turkey, and turtle–toad), each following this sequence: (1) comprehension assessment, (2) four questions on inheritance of properties, (3) control question, (4) question on kindhood judgement after growing-up, (5) introduction of blood transfusion scenario and (6) question on kindhood judgement comprehension question checked whether birth and adoptive parents are known; control question checked for exclusive birth bias	Design: experimental design (between-subject: community, age; within-subject: trait type and trait familiarity) Measurement: two questions on behavioural trait and two questions on physical trait (for each trait one familiar and one novel trait, counterbalanced); one question on kindhood judgement before and after transmission mechanism scenario	Main effect age: birth bias increased with age (4–10 yrs; p < .05) Main effect trait type: birth bias was stronger for physical trait type (vs. behavioural; p < .01) Main effect familiarity: birth bias was higher for familiar traits (vs. novel; p < .001) Interaction effect trait type × familiarity × age: birth bias became more pronounced for physical (vs. behavioural; p < .001) and for familiar (vs. novel; p < .05) with age Kindhood attributed to birth parents (birth parents bias; ps < .05) Main effect community (for transmission mechanism), Native American children tend to judge that blood transfusion affects kindhood (p < .001)	Children develop biological component in reasoning about trait inheritance and kindhood attribution in early childhood Differentiated pattern develops at the age of 9–10 yrs
Waxman et al. (2007) – Experiment 2	4–5 yrs; n1 = 58; M = n/a; n2 = 103; M = n/a 6–7 yrs; n1 = 106; M = n/a; n2 = 128; M = n/a 9–10 yrs; n1 = 107; M = n/a; n2 = 110; M = n/a adult; n1 = 77. M = n/a; n2 = 78; M = n/a	Procedure was split into two sessions: first session comprised one adoption scenario and steps (1–4), second session used two different adoption scenarios (rabbit–raccoon and deer–sheep) with introduction of a transmission scenario (nurture or blood transfusion, counterbalanced)	Design: experimental design (between-subject: community, age; within-subject: question type, trait type and trait familiarity) Measurement: same as above, but see transmission scenario	Replicates the findings of experiment 1 and extends it to different communities (suburban and urban) and question type (blood vs. nurture) Main effect age: birth bias increases with age (4–10 yrs, p < .05) Main effect question type: birth-parent bias for nurture vignette (p < .001) Interaction effect question type × community (p < .01): less birth parent choices in case of blood transfusion for Native American children (p < .05)	Younger children more flexibly attribute birth-parent kindhood because they search for a biological essence The transfer mechanism of this essence is influenced by cultural discourse (such as blood in Native American culture)

Note. yrs = years; mos = months; n/a = not available; n.s. = not significant; OR = Odd’s ratio; IRR = inter-rater-reliability; BMDR = birth mother detection rate

The analysed articles presented investigations on three aspects of understanding inheritance, separately or in combination: (1) where do traits originate from; (2) how are traits passed on to offspring; and (3) what degrees of relatedness exist? Six investigations used an experimental study design varying personal variables (i.e. age or community) and/or context variables such as trait types (i.e. behavioural, physical), and trait familiarity (i.e. familiar, non-familiar). The outcome variables focused mainly on judgements of trait origins (i.e. birth or rearing/adoptive parent matches), explanations concerning the mechanisms of trait transfers (biological vs. non-biological mechanisms), and/or children’s concepts of family and relatedness.

According to Waxman et al. (2007) the tendency to attribute kindhood based on the biological parents emerges in early childhood and is independent of cultural influences. This is interpreted as evidence for the belief in biological mechanisms that fix kindhood at birth and keep it stable even in cases of environmental change (such as adoption). At the age of 11 through 14 years, the birth-parent bias stabilises when physical traits are more often attributed to birth parents than behavioural traits. However, younger children more flexibly attribute traits to birth or adoptive parents when confronted with transfer mechanisms (blood transfusion from or nurture by adoptive parents). Children from a culture in which blood plays a central role less often made birth-parent choices in the case of blood transfusion scenarios. This indicates that children account for those mechanisms depending on the transfer of biological essence and culture.

In contrast, Williams and Smith (2010) investigated three concepts of inheritance that had been considered separately before (i.e. origin of features, causal mechanisms of inheritance, and family and relatedness) in order to establish a developmental sequence of these concepts and to examine whether understanding of these concepts is context dependent. The authors found that children show a step-wise development of inheritance concepts (i.e. concepts are not acquired simultaneously). Young children (4 years old) seem to have implicit concepts of the biological inheritance of physical features without explicit understanding of the relevant biological mechanisms. However, by the age of 7 years, explicit inheritance concepts emerge and children incorporate specific knowledge concerning biological mechanisms. The last concept (relatedness and kinship) is finally mastered during adolescence (11–14 years old). However, neither children nor adolescents showed consistency in inheritance explanations across different tasks.

In a follow-up study, Williams (2012) found another step-wise development in children’s performance concerning phenotypic similarity and phenotypic difference. Young children (4 years old) were able to make basic inheritance judgements, although the ability to verbalise explanations explicitly was delayed to older than 7 years. Moreover, the concept of phenotypic similarity seemed to be acquired earlier than knowledge of phenotypic difference. However, in concordance with Williams and Smith (2010), the children lacked consistency in inheritance explanations within and between the tasks presented. The author reasoned that for children one problem with understanding inheritance may be the fact that the biological process of inheritance is not experienced immediately.

Another study that focused on children’s understanding of biological inheritance through detecting kinship (i.e. the biological mother of a newborn) was carried out by Kaminski et al. (2013). The authors realised that young children (5 and 7 year olds) were not capable of identifying birth mothers successfully, while 9-year-old children showed a consistent ability to match a newborn’s face with its mother, but only when no distracter comparison stimuli were shown. Younger children often lack differential weighting of perceptual features. Indeed, older children (11 years old) and adults are more capable of matching newborns and their mothers correctly when ignoring perceptual non-relevant information. Nevertheless, perceptual detection of kinship is complex and even adults did not solve these tasks perfectly (although at a rate above that of chance; Kaminski et al., 2013).

In an exploratory case study, Ergazaki et al. (2014) examined whether young children (4–6 years old) recognise inheritance patterns (i.e. physical traits vs. behavioural traits) and whether physical traits are explained by a pre-biological model involving the contribution of both parents and genes. Based on their results, children have conceptual difficulties in explaining the origin of both physical and behavioural traits. In fact, they underestimated the impact of social environment on behavioural traits, while overestimating its role on physical appearance. However, building upon children’s essentialist bias by using genes as a ‘conceptual placeholder’ (Solomon & Johnson, 2000, p. 88), a precursory model of inheritance can be promoted, at least for physical traits (Ergazaki et al., 2014).

In a follow-up study, Ergazaki et al. (2015) examined children’s (5 years old) reasoning about reproduction and kinship. During a teaching intervention, the authors introduced a rudimentary idea of genes as a possibility for children to explain the species and body trait resemblance better. For this, they divided the interventions into sessions concerning ‘species-genes’ and ‘body-trait-genes’ (Ergazaki et al., 2015, p. 3136). The first term explains that a couple from one species (e.g. rabbit mum and dad) can only have babies of their own species (i.e. baby-rabbits, not puppies). In contrast, the second term enables children to offer birth-driven arguments for the body traits of offspring in contrast to nature-driven behavioural features (birth-parents vs. adoptive-parents; Ergazaki et al., 2015; Ergazaki, 2018). Based on a mixed-model case study with individual semi-structured interviews before and after a teaching intervention, the authors discovered that their teaching intervention was effective in several ways. Firstly, children’s reasoning about (1) reproduction, (2) species resemblance, and (3) body trait resemblance increased from pre- to post-interviews. Secondly, children showed an increased consistency in desired claims and desired justifications by including the newly learned content in their post-interviews. The idea of ‘species-genes’ in contrast to ‘body-trait-genes’ was mostly used after the teaching intervention (Ergazaki et al., 2015, p. 3136). Thus, introducing young children to an ‘essence-like’ species gene inheritance idea seemed to facilitate their basic level understanding of the inheritance concept (Ergazaki et al., 2015, p. 3136).

Recently, Raman (2018) focused on children’s understanding of inheritance of two biological processes (i.e. height and weight) and changes in this understanding when nutrition information (i.e. healthy or unhealthy food) was given. The author found a developmental difference, with younger children (4–5.5 years old) attributing less correct connections to height than adults did. In particular, third graders (8–10 years old) seemed to identify inheritance as a more relevant factor in the determination of weight than the influence of nutrition, while this idea declined in fifth graders (10–11 years old) and adults. However, children (and adults) viewed height as a more inherited, stable attribute (i.e. height is not strongly influenced by environmental factors such as nutrition). Thus, children might have a rudimentary framework for recognising that biological processes such as height (in contrast to weight) are more influenced by genetics than by nutrition. However, it is highly unlikely that children have the same kind of understanding of the underlying biological mechanisms as adults do.

In contrast to the previous investigations, Moya et al. (2015) analysed whether people of different nations (i.e. Puno, Peru: 4–75 years old; Yasawa, Fiji: 5–73 years old; United States: 18–64 years old) differentiate between (1) cultural and pre-natal transmission of traits as well as (2) parental and non-parental social influences. The results indicated that younger participants (4–10 years old) fail to differentiate the two kinds of traits in both adoption and migration condition tasks. Moreover, they seemed to have a slight birth bias for all traits (i.e. 60–80% chose a birth-parent resemblance). However, by middle to late childhood (approximately 11–15 years old), participants more often expected morphological traits to be under pre-natal influence, while belief traits were more culturally influenced. The authors concluded that ‘humans cross-culturally come to expect different effects of social and pre-natal influence, but develop culturally specific beliefs about the degree of parental social influence’ (Moya et al., 2015, p. 608).

Overall, the review of the studies focusing on inheritance revealed that even young children (4–6 years old) seem to have an implicit knowledge of inheritance. With increasing age, children (7–10 years old) were able to verbalise explicit explanations of inheritance (explicit knowledge). They also implemented genes within their frameworks if genes were introduced as ‘essence-like’ ‘species-genes’ (Ergazaki et al., 2015, p. 3136). However, children may not yet be able to understand the underlying genetic aspects as adults do. Moreover, cultural aspects can have an influence on the understanding of trait transmission (e.g. Moya et al., 2015).

Children’s understanding of natural selection

Four of the analysed articles focusing on the core concept selection (Emmons et al., 2016; Emmons et al., 2018; Kelemen et al., 2014; Shtulman & Harrington, 2016) used storybook interventions to explore children’s explanations of adaptation by natural selection (Table 4). A fifth article explored the impact of narrative language on children’s language use at recall, as well as their endorsement of explanations for change (Legare et al., 2013). The last article was concerned with children’s understanding of the origin of species and its relationships with natural history knowledge and parents’ beliefs (Evans, 2000). However, this study was nearly two decades old and, to be consistent with the decision taken about inheritance article publication dates, we excluded this article from further analysis.

Table 4. An overview of the empirical studies focusing on the core concept natural selection in early childhood evolution education

Study	Sample Age group; sample size; mean (M) age (SD/range)	Intervention/procedure	Design & measurement	Results	Conclusions
Emmons et al. (2018)	Kindergarteners; n = 20; M = 6 yrs 2 mos (SD = 4 mos) Second graders; n = 17; M = 8 yrs 5 mos (SD = 5 mos)	Two storybooks were presented because comparing examples promotes the near and far generalisation of biological concepts	Design: within-subject design that introduces the experimental variation in the post-test questions (questions on near AND far transfer/generalisation) Measurement: comprehension, near and far transfer and delayed test on natural selection (κ = 0.89)	Kindergarteners: comprehension pre–post increase in factual knowledge by OR = 11.68 near generalisation on book 1 indicate stable and transferrable factual knowledge (n.s.) and on book 2 even increased understanding of NS (OR = 2.37) far generalisation on book 2 led to performance decline by OR = 0.26 with even less factual knowledge articulated delayed testing indicated neither increases nor further decreases on near and far transfer (both n.s.) Second graders: comprehension pre–post increased by OR = 120.74, No further changes in understanding on near and far generalisation or delayed tests (all n.s.)	Kindergarteners’ understanding of natural selection is based on factual knowledge that is transferrable in near generalisation scenarios (foraging traits) Structural similar storybooks hindered far transfer in which even factual knowledge was not generalised pronounced knowledge of camouflage and within-species variation enhanced Understanding of natural selection in second graders
Emmons, Smith & Kelemen (2016)	5–6 yrs; n = 16; M = 6 yrs 2 mos (SD = 6 mos) 7–8 yrs; n = 16; M = 8 yrs 3 mos (SD = 3 mos)	Same as in Kelemen et al. (2014), but under field conditions (e.g. distractors such as noise)	Design: field experiment with storybook intervention and two groups (5–6-year-olds; 7–8 year-olds); pre- and two post-tests (clinical interview with six closed- and four open-ended questions) Measurement: comprehension and generalisation of concepts (κ = 0.97)	Comprehension: 5−6-year-olds: pre–post 1: OR = 22.57 (increased) 7−8-year-olds: pre–post 1: OR = 66.36 (increased) Generalisation: 5−6-year-olds: pre–post 2: decrease by OR = 0.30, but n.s. 7−8-year-olds: pre–post 2: n.s.	Older children outperformed younger children: younger children acquired and generalised predominantly isolated facts, however older children’s abilities were more pronounced This may be because older children not only have higher cognitive and linguistic skills but also a better representation of biological variation (see also Emmons & Kelemen, 2015)
Shtulman et al. (2016)	4−6 yrs; n = 52; M = 5;6 yrs (4;0–6;9) 7−12 yrs; n = 44; M = 5;6 yrs (7;0–12;8) adults; n = 30; M = n/a (18−22)	Picture book with two items each for pre-test, training, post-test	Design: experimental study with pre–post-test design and intervention on three groups (4–6-year-olds; 7–10-year-olds 2; adults) Measurement: coding of evolutionary explanations and alternative conceptions	Interaction effect assessment × age (evolution score): 4−6-year-olds: pre–post: d = 0.55; 7−12-year-olds: pre–post: d = 1.19; adults: pre–post: d = 1.13 Main effect age (alternative conceptions): fewer in young children, similar in older children and adults	Older children are capable of learning evolutionary principles Analogical encoding is a possible teaching method High cognitive load for younger children/other conceptions in the first place
Kelemen et al. (2014) – Experiment 1	5–6 yrs; n = 28; M = 5;9 yrs (SD = 6 mos) 7–8 yrs; n = 33; M = 7;9 yrs (SD = 5 mos)	Storybook with realistic pictures and factual narrative of evolutionary explanations on six key concepts (no teleological language used)	Design: quasi-experimental study with storybook intervention and two groups (5–6-year-olds; 7–8-year-olds); pre–post- and follow-up test Measurement: clinical interview with five closed and five open-ended questions on comprehension and generalisation of natural selection (day 1); long-term comprehension and generalisation of natural selection (day 2); coding rubric with five levels (κ = 0.84)	Comprehension: 5−6-year-olds: pre–post: OR = 18.86, follow-up: n.s. 7−8-year-olds: pre–post: OR = 11.54, follow-up: n.s.	Children’s understanding of natural selection is coherent, transferable and sustainable, however, more pronounced amongst 7−8-year-olds Limited to change in initial population
Kelemen et al. (2014) – Experiment 2	5–6 yrs; n = 16; M = 6;0 yrs (SD = 4 mos) 7–8 yrs; n = 16; M = 8;3 yrs (SD = 3 mos)	Same as exp. 1, but more prompts for self-explanation	Design: same as exp. 1 but without follow-up test Measurement: same as exp. 1, but κ = 0.89	Comprehension: 5−6-year-olds: pre–post: OR = 42.17 7−8-year-olds: pre–post: OR = 38.98	Replicated and extended findings of exp. 1 Children understand nuances of natural selection Older children can transfer natural selection to novel species by more detailed storybook
Legare et al. (2013)	5.5–8 yrs; n = 34; M = n/a (5.5–7.9 yrs) 8–12 yrs; n = 54; M = n/a (8.0–12.4 yrs)	Three narrative types randomly ordered (desire-, need- or selection-based) on five key concepts	Design: quasi-experimental design with two factors: narrative type (within) and age group (between) Measurement: structured interviews with closed and open-ended questions after each narrative (three tests)	Main effect age: older children memorised more details (location: p < .05; ancestors: p < .01) and gave more need-based explanations (p < .01) Main effect narrative type: narrative type prompts explanations (desire-base: ps < .001; need-based: ps < .01; selection-based: ps < .07)	Need-based reasoning is essential in naïve biology and therefore scaffolds evolutionary reasoning

Note. yrs = years; mos = months; n/a = not available; n.s. = not significant; OR = Odd’s ratio

Legare et al. (2013) showed that the narrative language influences the recall of explanations and concepts of evolution as well as the language used by children (e.g. children used anthropomorphic explanations and language after they heard an anthropomorphic narrative). In a quasi-experimental study, the authors varied age group (between-subject: 5–7 and 8–12 years old) and narrative type (within-subject: desire-, need- or selection-based language) in order to investigate the influence of language on children’s recall of evolutionary explanations. Measurement of the children’s recall of explanations from the narratives revealed the main effects of age group and narrative type: both need-based and selection-based narratives prompted the use of more evolution concepts in their explanations. Older children (8–12 years old) were more likely to recall concepts of within-species variation and the inheritance of traits, as well as differential survival and reproduction, than younger children (5–7 years old). Legare et al. (2013) concluded that need-based narratives provide a stepping stone to understanding evolution for younger children (5–7 years old).

In a study by Kelemen et al. (2014), children aged 5 through 6 and 7 through 8 years understood and generalised the concept of natural selection after a storybook intervention. The findings were extended from a comprehension of natural selection in one initial generation (experiment 1) to natural selection over multiple generations (experiment 2). In a quasi-experimental study, two groups of children were presented with a storybook intervention and tested with pre-, post-, and follow-up assessments. The assessments probed the children’s factual knowledge with closed questions as well as self-generated explanations with open-ended questions. The storybook used non-intentional and non-teleological language for six mechanistic explanations of natural selection: (1) within-species variation, (2) ecological changes, (3) differential survival and (4) reproduction, (5) trait inheritance, and (6) trait-frequency change. One limitation reported by the authors was that the inherited traits were function- and foraging-related. Furthermore, the intervention was applied in a laboratory situation and implied extensive scaffolding (Shtulman & Harrington, 2016).

Emmons et al. (2016) replicated the findings by Kelemen et al. (2014). However, they applied the storybook intervention in a more realistic field setting, in an after-school programme, with distractors such as other visual and learning materials present in the room and a moderate noise level outside the room. After the storybook intervention, for 5 through6 and 7 through8-year-old children from a more racially diverse sample than Kelemen et al. (2014), the older children outperformed the younger. While the younger children acquired and generalised isolated facts, the older children gave more pronounced explanations. The authors concluded that older children have not only higher cognitive and linguistic skills but also a better understanding of within-species variation (Emmons et al., 2016; see also Emmons & Kelemen, 2015; Shtulman & Schulz, 2008).

In contrast, the article of Shtulman and Harrington (2016) compared children’s (4−6 and 7−12 years old) and adults’ (18−22 years old) ability to explain biological adaptation by natural selection. They built their study on the findings of Legare et al. (2013), that children recall concepts of evolution when non-anthropomorphic language is used, and the findings of Kelemen et al. (2014), that children are able to self-generate explanations for natural selection. In addition, the authors addressed four limitations of the Kelemen et al. (2014) study, by: (1) including adults as a comparison group; (2) comparing two types of traits, foraging- and camouflage-related traits, that were also function-related; (3) coding children’s explanations analytically (i.e. for evolutionary principles) rather than holistically (i.e. for logical coherence of facts); and (4) using a single prompt. Compared with adults, only older children (7−12 years old) showed similar learning gains. Moreover, both foraging- and camouflage-related traits were equally mastered. Thus, the older children (as well as adults) used different evolutionary principles in their explanations. The authors also suggested that the use of multiple examples for different traits (i.e. foraging and camouflage) promoted ‘analogical encoding’ (Shtulman & Harrington, 2016, p. 1225).

Recently, Emmons et al. (2018) reported on kindergarteners’ and second graders’ abilities for ‘near and far transfer’ (p. 321) of mechanistic explanations on natural selection not only to structurally similar foraging-related traits but also to camouflage-related traits. Two storybooks with mechanistic explanations of natural selection of foraging-related traits were used, because comparing examples enhances the near and far transfer of biological concepts (see also Shtulman & Harrington, 2016). Unlike Shtulman and Harrington (2016), Emmons et al. (2018) coding of children’s answers on near and far generalisation, as well as delayed tests, accounted for factual knowledge (level 1a+1b) and coherent explanations (level 2−4). Kindergarteners’ understanding of natural selection was mostly the result of increased factual knowledge (based on a comprehension test after reading book 1) and transferrable to other foraging-related traits. However, analogical encoding from book 2 promoted coherent explanations on near generalisation to another foraging-related trait (achieving at least level 2), but when tested for their far generalisation ability on a camouflage-related trait, kindergarteners’ performance dropped and even factual knowledge was no longer available (60% at level 0). One-month delayed testing on near and far transfer indicated no further changes. With second graders and the same two storybooks, there was an almost 120-fold increase in factual knowledge and explanations on natural selection, even after reading book 1. Near and far generalisation led to no further increases. A simple explanation for the kindergarteners’ difficulties with far transfer is a lack of knowledge on the concept of camouflage (which was present for the second graders). Another reason for the limited understanding of natural selection, as reinforced by the results, is kindergarteners’ more pronounced essentialist bias (compared with second graders) that interprets trait changes as a result of age-related growth at an individual level and not a population level (Emmons et al., 2016; Emmons & Kelemen, 2015).

In summary, the review of the five studies concerning children’s abilities to understand natural selection revealed that children (5−8 years old) recall and self-generate explanations for natural selection by drawing on evolutionary concepts such as variation, differential survival and reproduction, inheritance, and population change. However, these findings are moderated by age as well as intervention measures. Older children (7–12 years old) used more pronounced explanations for natural selection that were not only based on factual knowledge, but on coherent explanations, and were more able to generalise their understanding to different trait types than younger children (4–6 years old; Emmons et al., 2016; Shtulman & Harrington, 2016). One reason for a more coherent understanding of natural selection is the enhanced representation of within-species variation in older children (Emmons et al., 2016) and a less pronounced essentialist bias of trait change at an individual level (Shtulman & Harrington, 2016). However, it is still open to debate how structural similarity of learning materials (e.g. story and picture books) promotes the generalisation of facts and explanations of natural selection in kindergarteners. While older children did not have any problems with analogous encoding, structural similarity (Emmons et al., 2016) as well as increased dissimilarity (Shtulman & Harrington, 2016) did not promote far generalisation to trait types other than foraging-related traits. The question is how structural similarity is construed from an evolutionary point of view (i.e. foraging or camouflage; gain or loss; animal or plant) as several surface features influenced the explanations given for natural selection. More differentiated prompts (up to 10 forms) enhanced the use of evolutionary explanations and helped both younger and older children (Emmons et al., 2016; Kelemen et al., 2014), while minimal prompts (two forms) were not effective (Shtulman & Harrington, 2016).

Discussion

This review provides a summary of the evidence available for children’s precursory concepts concerning evolution as well as cognitive biases and how they potentially affect learning about evolution in formal schooling (e.g. in primary school). Numerous studies have identified difficulties in learning about evolution throughout education (e.g. Banet & Ayuso, 2003; Bishop & Anderson, 1990; Harms & Reiss, 2019; Lawson & Thompson, 1988). There is evidence that some of these difficulties stem from cognitive biases that develop early, before schooling (Opfer et al., 2012; Shtulman & Calabi, 2012). According to Shtulman and Harrington (2016), until 2016 only two studies (Kelemen et al., 2014; Legare et al., 2013) had investigated children’s understanding of concepts in evolution. From our review, it is clear that an increasing number of published studies are investigating children’s understanding of evolutionary core concepts such as variation, inheritance, and selection. While research on children’s understanding of variation and natural selection is quite young (since 2008), research on children’s understanding of inheritance dates back to a study by Springer and Keil (1989). Our analysis shows that it is not enough to consider research results on children’s precursory concepts of only natural selection (e.g. Shtulman & Harrington, 2016) if the goal is to identify what children already understand about evolution. The contribution of evolutionary core concepts such as variation and inheritance on children’s understanding of basic evolutionary mechanisms has been explored (e.g. Emmons & Kelemen, 2015; Ergazaki et al., 2014). Evolutionary core concepts (i.e. variation, inheritance, and selection) facilitate information encoding and long-term recall (Inagaki & Hatano, 2004; Opfer et al., 2012; Vosniadou & Brewer, 1992), and thus we discuss below the educational implications of how each of these core concepts contributes to children’s understanding of basic evolutionary mechanisms.

Children’s understanding of variation

Consistent with previous literature on learners’ struggles with individual variation (e.g. Bishop & Anderson, 1990) and the development of children’s essentialist bias (e.g. Gelman & Coley, 1990), research shows that children need knowledge of within-species variation to understand basic mechanisms of natural selection (Emmons et al., 2016; Emmons & Kelemen, 2015). Children’s awareness of within-species variation is central to learning about natural selection, and even older children do not consider individual differences within a species as a condition that precedes the emergence of a trait through natural selection (Bishop & Anderson, 1990; Ferrari & Chi, 1998). However, essentialist reasoning prevents children from observing within-species differences (Coley & Muratore, 2012; Shtulman & Calabi, 2013) because children’s assumptions about the animals’ category membership (i.e. the species) within the essentialist bias can outweigh the perception of animals’ surface characteristics (i.e. the phenotype; Gelman & Coley, 1990).

On the other hand, previous research has shown that learners’ beliefs are stepping stones towards understanding evolution in school (e.g. Zabel & Gropengiesser, 2011) and for younger learners (e.g. Evans et al., 2012). In particular, teleological beliefs are regarded as an intermediate step when considering children’s knowledge of within-species variation (Legare et al., 2013; Evans et al., 2012). Children’s teleological bias can promote their understanding that species’ traits can vary (‘are mutable’; Emmons & Kelemen, 2015, p. 158) when useful traits are acquired (Legare et al., 2013). Hence, children’s teleological attribution of traits that are based on individuals’ needs might help them reconsider their essentialist bias of trait stability within a species because of a species’ essence. As suggested previously (e.g. Zohar & Ginossar, 1998), children’s teleological bias can be built upon by emphasising structure–function relations through need-based language in instructional materials (Legare et al., 2013). Hence, need-based reasoning on structure–function relations are also an intermediate step in learning about evolution.

Nevertheless, a focus on trait function can hinder the learning of natural selection as a variation-based process, because trait function may emphasise need-based variation (Emmons & Kelemen, 2015). Here, both teleological and essentialist bias mutually influence each other (‘coalescing’; Emmons et al., 2016, p. 1206): the children’s assumptions about a trait function related to the need for survival promote essentialist assumptions about the stability of the trait. Indeed, including trait function in stimuli materials (i.e. depictions of animals’ phenotypes) decreased the variation acceptance rate of 7- to 8-year-old children (study 1) whereas omitting trait function increased the acceptance rate to 60% (study 2; Emmons & Kelemen, 2015). Proposals for using need-based reasoning to enhance the understanding of mutability of species can have unwanted long-term consequences (Emmons & Kelemen, 2015; cf. Legare et al., 2013) and it can impede the transfer of evolutionary processes (e.g. variation) to other contexts through analogical encoding (Shtulman & Harrington, 2016). As already known from research on evolution assessments, context (e.g. trait gain vs. loss) affects students’ explanations of biological change through evolution (Nehm & Ha, 2011). Care needs to be taken as need-based reasoning is especially present in reasoning on the loss of traits (Nehm & Ha, 2011). More research is needed to investigate whether focus on function promotes misconceptions in early evolution learning.

Children’s understanding of inheritance

Inheritance cannot be experienced immediately, yet the analysed studies make it clear that even children as young as 4 through 6 years have an implicit understanding of the inheritance concept. However, some caution is required when talking about children’s understanding of inheritance, because young children are unlikely to have the same understanding of the underlying biological mechanisms as adults do (e.g. Raman, 2018; Williams, 2012). Children’s personal interest in nature and their essentialist-like assumptions can be used as a stepping stone in the learning progression (e.g. Duschl et al., 2011) for teaching inheritance. Thus, an essence-like species-genes idea can be used as a placeholder to support children’s understanding of inheritance, although this has to be done at a very basic level and with care to reveal their false essentialist assumptions (e.g. Ergazaki et al., 2014, 2015). Interestingly, the role of the inheritance concept in understanding evolution has not been investigated in early ages, while articles on other concepts (e.g. natural selection: Emmons et al., 2018) emphasise the relevance of inheritance for understanding evolutionary processes. The influence of ‘essence-like’ ‘species-genes’ (Ergazaki et al., 2015, p. 3136) as a conceptual placeholder for inheritance in understanding variation would be of interest as essentialist assumptions are known to hinder the understanding of within-species variation. Again, because of their teleological assumptions, children more easily assume traits providing a function are heritable (Ware & Gelman, 2014).

Children’s understanding of natural selection

Around the age of 7 through 8 years, children exhibit a more pronounced understanding of natural selection by not only possessing factual knowledge but also providing coherent explanations (Emmons et al., 2018; Shtulman & Harrington, 2016). Hence, teaching natural selection in primary schools seems achievable. Early learning of core concepts in evolution may help learners build a deeper understanding and prevent cognitive biases such as teleological and essentialist assumptions from negatively influencing each other. At the same age (7–8 years old), children’s teleological assumptions on a trait’s relevance to survival promote their essentialist assumption that survival-relevant traits will not vary between individuals within a species (Emmons et al., 2016; Emmons & Kelemen, 2015; Shtulman & Schulz, 2008). In older age groups, essentialist and teleological biases in particular are more pronounced and related to students’ reasoning on natural selection (Opfer et al., 2012). However, evidence is still missing on how to support children younger than 7 years. It seems that representing within-species variation is essential (and more pronounced with older children) for understanding natural selection (Emmons et al., 2016; Emmons & Kelemen, 2015; Legare et al., 2013).

From a methodological point of view, studies differ in how children’s understanding of natural selection is described: coding children’s explanations analytically (Legare et al., 2013; Shtulman & Harrington, 2016) rather than holistically (Emmons et al., 2016; Emmons et al., 2018; Kelemen et al., 2014) provided further insights concerning which evolutionary principles were included in children’s explanations, but lacked the logical coherence of concepts used to describe natural selection (Emmons et al., 2018). However, as shown in research on the understanding of natural selection in older age groups, analytical coding schemes are able to detect the coexistence of normative and non-normative concepts of natural selection (Opfer et al., 2012).

Implications for early childhood educators and researchers

Although we cannot present a one-fits-all solution for how to implement evolutionary core concepts or related objects and ideas (e.g. fossils) in early childhood education, the current findings support the design of some of the available resources for childhood educators (for a comprehensive bibliography on teaching resources, see Russell & McGuigan, 2015). In particular, children’s understanding of within-species variation seems crucial and is acknowledged in learning progressions (e.g. Furtak, 2012; Lehrer & Schauble, 2012) as well as in several teaching materials (e.g. Anton et al., 2012; Campos & Sá-Pinto, 2013). Teaching resources rely on children’s interest in the natural world and try to use and foster their emerging scientific skills such as observation of their environment (e.g. Klemm & Neuhaus, 2017; Lehrer & Schauble, 2012). For instance, the citizen science project Evolution MegaLab facilitates the observation of shell polymorphism in banded snails (Worthington et al., 2012) and created materials that can be used to foster young children’s understanding of within-species variation (e.g. Anton et al., 2012). By helping children to observe and recognise differences between individuals of a species (e.g. shell polymorphism in banded snails), educators can support young children’s understanding of within-species variation, which is a prerequisite for natural selection to occur. With regard to children’s cognitive biases, the findings indicate that everyday language may foster the emergence of cognitive biases and, thus, educators should keep this in mind. Hence, implementing pre-scientific ideas such as species-genes can help to prevent cognitive pitfalls (Ergazaki et al., 2015). These findings seem important to us, because even within secondary classrooms, both students and teachers might be unaware of teleological notions within explanations (Gresch & Martens, 2019).

Based on our literature review, only few studies considered how interventions on evolution have to be designed for children younger than 7 years to promote pre-scientific understanding of evolutionary core concepts (cf. Emmons et al., 2018). We see at least two possible further research strands that could investigate the interrelation of evolutionary core concepts in more detail (e.g. for variation and natural selection; Emmons & Kelemen, 2015) and broaden the view towards neighbouring precursory concepts (e.g. for fossils; Borgerding & Raven, 2018). First, research on early childhood education should further explore the core concepts of variation, inheritance, and selection, as well as children’s cognitive biases both alone and in combination with each other (for undergraduate students, see Opfer et al., 2012). For instance, despite the great number of studies that focused on the inheritance concept, it was seldom investigated in the context of natural selection or in combination with the variation concept. Furthermore, following the contradictory evidence on the usefulness of cognitive biases when children have already developed scientifically correct precursory concepts, it has been suggested that it is better to build upon the latter rather than the former (Emmons & Kelemen, 2015). However, as cognitive biases and normative ideas co-exist at least in older students (Opfer et al., 2012), we suggest that further research explores how this co-existence develops in early childhood. Second, researchers should also explore in more detail how other precursory concepts such as children’s understanding of food webs (e.g. how organisms rely on each other for nourishment; Allen, 2017) or their notion of living/non-living may influence their understanding of precursory core concepts of evolution (see also section Core concepts in evolution within science standards). Future studies could, in a next step, investigate the context-dependency of children’s precursor concepts of evolution by not only focusing on core concepts in the evolution of animal species (e.g. Emmons et al., 2018) but also in the evolution of plant species. It would be interesting to see further research on whether context-dependent misconceptions (e.g. depending on the taxa of plants and animals; Nehm & Ha, 2011; Opfer et al., 2012) are rooted in children’s tendency to not attribute plants to living things before the ages of 7 through 9 (Hatano & Inagaki, 1994). This tendency is present even in adults (Goldberg & Thompson-Schill, 2009). At this point, a literature review on the impact of children’s notions of living/non-living on precursory concepts of evolution is missing.

Limitations

Literature reviews are generally limited to the literature that is available on the databases as well as being limited by the search and selection procedure. We chose to limit our analyses to peer-reviewed articles published in research journals and indexed in any of the three databases used (i.e. ERIC, PsychINFO, and Web of Science). We are aware that this decision may have limited the available literature (i.e. if articles are not indexed in the respective databases) and forced us to exclude certain publications (e.g. book chapters, research reports) that may have provided valuable insights into children’s understanding of precursory concepts in evolution. However, searching in three databases that index some of the most visible journals, we aimed to find the majority of important contributions to the field of our research. Moreover, our focus on peer-reviewed journal articles ensured a certain level of research quality by relying on journals’ policies concerning the review process.

Another limiting aspect of our review is the selection of keywords for the literature search. For our search, we limited our keywords to the core (key) concepts of variation, inheritance, and (natural) selection, which are the scientifically normative causal ideas for explaining evolutionary change (Opfer et al., 2012). Including other key (e.g. limited resources, change over time, speciation; Anderson et al., 2002; Nehm & Schonfeld, 2008) or threshold concepts (i.e. randomness, probability, temporal scales/deep time, and spatial scales; Tibell & Harms, 2017) could have resulted in additional articles. For instance, understanding temporal scales is certainly foundational for making sense of evolution, since ‘evolution includes processes that occur over timescales ranging from extremely short time for mutations (submilliseconds) to deep time (millions of years) for macroevolution of species and higher taxa’ (Göransson et al., 2020, p. 4). In fact, learners of all ages (children, adolescents, adults) have difficulties in understanding processes or events on temporal scales that go beyond people’s sensory experience (e.g. Cheek et al., 2018; Piaget, 1969). However, empirical studies on deep time in the context of evolution often focus on factual knowledge in which learners should correctly determine the time or timing of particular events (e.g. emergence of photosynthesis, age of earth; Catley & Novick, 2009; Cheek, 2012; Hidalgo et al., 2007; Johnson et al., 2014). Thus, we decided to focus on the three main core concepts of evolution, also considering the fact that additional key concepts could be amalgamated under these aspects. In addition, the notion of threshold concepts in evolution is fairly new, so that we cannot yet expect results in young children.² Finally, the research field on the precursor concepts of evolution in early childhood is rapidly growing and new valuable articles on young children’s understanding of evolutionary core concepts may have appeared since the initial research for this literature review was conducted (e.g. Borgerding & Kaya, 2019).

Conclusion

The analysed studies reveal that even young children aged up to 7 years are capable of understanding evolutionary core concepts such as variation, inheritance, and selection. While children’s understanding of the inheritance concept has been intensively investigated (Table 3), children’s understanding of variation has not been considered in detail (Table 2). This is critical as understanding variation is an indispensable pre-requisite for understanding evolutionary change in the long term. In particular, the dependency of children’s precursory concepts of natural selection on their ability to represent variation in species provides an interesting avenue for further research (e.g. Emmons et al., 2016). Variation and related underlying concepts such as randomness are often highlighted as a major obstacle in learning evolution (e.g. Batzli et al., 2016; Fiedler et al., 2017; Garvin-Doxas & Klymkowsky, 2008). Thus, more research is needed on how to introduce younger children in particular to the concept of variation to promote evolutionary ideas and prevent essentialist assumptions.

Focusing on the use of children’s biases for further learning presents contradictory evidence. On the one hand, cognitive biases can serve as stepping stones in a learning progression (e.g. Duschl et al., 2011) for evolutionary principles (e.g. Ergazaki et al., 2015; Legare et al., 2013), but on the other hand, the usefulness of cognitive biases is questioned as children develop scientifically correct explanations (e.g. Emmons et al., 2016; Emmons et al., 2018; Kelemen et al., 2014). When teaching evolution is oriented towards learning progressions (e.g. Evans et al., 2012; Lehrer & Schauble, 2012), both essentialist and teleological bias are regarded as constraints that can either promote or hinder the learning of evolutionary concepts. Lehrer and Schauble (2012) argue that observing between- and within-species variation is a ‘building block’ (p. 704) in the learning progression on evolutionary thinking. However, children’s assumption about purposeful-designed traits (teleological) may ultimately lead to attribution of the respective trait to all individuals of the species (i.e. consolidation with essentialist assumption; Emmons et al., 2016), hence neglecting within-species variation. In contrast, teleological assumptions on adaptation (i.e. function of traits) strengthen the acceptance that traits vary (Evans et al., 2012). Furthermore, essentialist assumptions may provide a basis for children to acknowledge that individuals have an innate potential to change. Hence, children’s essentialist assumptions may be a stepping stone in a learning progression to understand common descent (Evans et al., 2012).

However, when teaching concepts of evolution, the learning context and language used (such as need-based language) must also be considered carefully (e.g. Emmons & Kelemen, 2015; Legare et al., 2013). Using well-designed learning opportunities can support children in early childhood to achieve their potential to understand the natural world (e.g. Ergazaki et al., 2015). In particular, children’s understanding of within-species variation can be supported by early experiences of different types of change that occur in living animals (e.g. Herrmann et al., 2013).

Acknowledgments

We thank Denise Bock for her support in the review process of individual articles for our data analysis. We are also grateful to John Blackwell for language review and the anonymous reviewers for their feedback and valuable suggestions on an earlier version of this manuscript.

Disclosure statement

No confliction of interest is declared by the authors.

Additional information

Notes on contributors

Till Bruckermann

Till Bruckermann is a post-doctoral researcher in the Department of Biology Education at the IPN – Leibniz Institute for Science and Mathematics Education in Kiel (Germany). His main research interest is scientific reasoning and inquiry-based learning with a focus on citizen science. He also engages in research on teaching and learning evolution.

Daniela Fiedler

Daniela Fiedler is a post-doctoral researcher in the Department of Biology Education at the IPN – Leibniz Institute for Science and Mathematics Education in Kiel (Germany). Her main research interests include teaching, learning, and accepting biological evolution across the educational stages.

Ute Harms

Ute Harms is director at the IPN – Leibniz Institute for Science and Mathematics Education in Kiel (Germany) and Professor for Biology Education at the University of Kiel. Her main research interests are conceptual learning in the life sciences, climate literacy, as well as biology teacher education, biology-related competitions, and transfer of contemporary topics in life sciences to the public.

Notes

1. Evolutionary biologists can show that changes (characteristics) acquired during an individual’s lifetime can be passed on to its offspring as a result of epigenetic inheritance (e.g. Hernando-Herraez et al., 2013; Nätt et al., 2012; Turck & Coupland, 2014). However, there is still some debate about the impact of epigenetics on evolution in the long term (e.g. Dickins & Rahman, 2012; Pennisi, 2013). Thus, for our description of the inheritance concept, we ignore epigenetic aspects, as children mostly misunderstand the inheritance concept because of their need-based assumptions rather than the differentiation of biological mechanisms. However, in later education teaching about ‘genetic material’ instead of ‘genes’ may help to distinguish genetic and epigenetic inheritance (Stern & Kampourakis, 2017, p. 214).

2. We are aware that research on the understanding of randomness and/or probability in young children does exist (e.g. Jones et al., 1997; Schlottmann, 2001), although these works are situated in the context of mathematics and not connected to the context of evolution. This is of relevance because conceptual knowledge of randomness and probability in the context of evolution differs from knowledge in the context of mathematics, at least for university students (Fiedler et al., 2019, 2017).