The development of early scientific literacy gaps in kindergarten children

ABSTRACT

In times of climate change, worldwide diseases, and the question of sustainable energy, scientific literacy has never been more important. The acquisition of scientific literacy starts in early childhood and depends on children’s experiences in learning environments such as the parental home and kindergarten. The Starting Cohort 2 of the German National Educational Panel Study offers longitudinal data on the scientific literacy of 2,937 children from kindergarten to the third grade of primary school. We used linear latent growth curve models to analyse the data. The results show that the scientific literacy of kindergarten children grew over time and that kindergarten children already differed in their scientific literacy. Especially children from non-German-speaking homes or whose parents had lower levels of education or who came from homes with low cultural capital had a significantly lower level of scientific literacy in kindergarten. A scientific focus in kindergarten, on the other hand, had a positive effect on the scientific literacy of kindergarten children. Moreover, our study did not reveal any interindividual differences in the growth of scientific literacy over time. Thus, our study did not find any evidence to indicate that school had closed the initial gap in children’s scientific literacy.

KEYWORDS:

• Scientific literacy
• developmental research
• early years/early childhood
• learning environments

Introduction

The relevance of science has increased enormously in the last years. The world is changing rapidly due to scientific and technological progress, which can be seen in environmental challenges or technological innovations, and scientific literacy (SL) is necessary to understand and deal with these changes (OECD, 2018). Imparting SL as early as possible not only enables individuals to deal with their daily problems, it also forms the basis for highly qualified individuals in science (OECD, 2018) and, ultimately, raises the economic growth of a country (Hanushek & Wößmann, 2015).

The acquisition of SL starts in early childhood and increases in line with cognitive and linguistic development in experiential environments (Gelman & Brenneman, 2004). Results from European studies as well as from assessments in the United States have consistently shown that there are two important learning environments in early childhood: the parental home and kindergarten (Melhuish et al., 2008; NICHD, 2002, 2003; Sylva et al., 2010; Tietze et al., 1998). These environments not only provide the framework for development, they also offer learning opportunities and materials, activities, and support for young children to promote their social, cognitive, and emotional development. However, due to the diversity of these environments, children naturally differ in their social background or the number of learning opportunities they have. Hence, it is not surprising that these differences can cause early disparities. Morgan et al. (2016) examined the SL of 7,757 preschool children in the United States and its development as the children progressed to school. They found significant performance differences between children in the fields of natural and social sciences when they entered kindergarten, which persisted beyond primary school. Children with a lower socioeconomic status (SES) or a language other than English spoken at home showed significantly lower science achievement than children with a high SES or children who spoke English at home (Morgan et al., 2016). Although some groups of children were able to compensate for this deficit, they were not able to fully close the initial gap in their science achievement (Morgan et al., 2016). Thus, the results of their study showed – in line with other studies – that early disparities in SL (Hahn & Schöps, 2019; Martin et al., 2012; Morgan et al., 2016; Saçkes et al., 2011), language skills (Becker, 2011; Becker et al., 2013; Ebert et al., 2013; Melhuish et al., 2008; Sammons et al., 2009), or numeracy skills (Anders et al., 2012; Klibanoff et al., 2006; LeFevre et al., 2009; Melhuish et al., 2008; Sammons et al., 2009) already exist in kindergarten and have an impact on children’s later life.

Despite the importance of science education and the acquisition of SL, there is a lack of research in Germany on the development of SL in young children and the possible factors that affect this development (e.g. in kindergarten). Cross-sectional studies such as the Trends in International Mathematics and Science Study (TIMSS; Martin et al., 2012) or the Programme for International Student Assessment (PISA; OECD, 2018) examine children’s SL in primary or secondary school. They provide insights into the current level of knowledge in the samples examined. However, to investigate the development of SL, longitudinal studies are needed. The aim of this study was to close this research gap by using the longitudinal data from the Starting Cohort 2 (kindergarten) of the German National Educational Panel Study (NEPS). NEPS offers the opportunity to study the development of various aspects of education, for example, SL, over the lifespan. The NEPS data make it possible to examine whether the results of Morgan et al. (2016) can also be found in Germany and whether initial differences in SL, linked to early social or migration-related disparities, persist into primary school.

Scientific literacy in NEPS

SL is a widespread concept that is used in large-scale national and international studies such as PISA (OECD, 2006). It also forms the basis for the scientific framework that was used to construct all of the science tests in NEPS (Hahn et al., 2013). SL implies an understanding of scientific concepts and processes. This understanding enables people to deal with everyday scientific situations and helps them to participate in a society in which science and technology are of great importance.

Within SL, two components can be differentiated between. First, knowledge about science (KAS) specifies the understanding of scientific processes, methods, and enquiry and also comprises knowledge of the means and goals of science (Hahn et al., 2013; OECD, 2006). KAS enables people to formulate hypotheses about the relationships between variables, to test hypotheses by conducting experiments, and to interpret the results (Zimmerman, 2007). Studies have shown that even 4-year-olds can distinguish between simple theories and empirical evidence while generating hypotheses (Koerber et al., 2005; Sandoval et al., 2014; Zimmerman, 2007). In early primary school, children learn to test simple hypotheses themselves by selecting appropriate examinations (Sodian et al., 1991) and to distinguish a confounded from a controlled experiment within a choice task (Bullock et al., 2009). Thus, the development of KAS starts at kindergarten age and continues into adolescence and adulthood (Sodian & Bullock, 2008).

Second, knowledge of science (KOS) includes the understanding of scientific content and the knowledge of concepts, theories, and terms from different content areas such as biology or physics (Hahn et al., 2013; OECD, 2006). The knowledge of concepts develops from children’s experiential environments and can help them to understand relevant phenomena in the world (Gelman & Kalish, 2007). Everyday experiences support children to develop their knowledge of concepts in a gradual process, which can include different levels of conceptual understanding (Gelman & Kalish, 2007). For example, previous studies established that although 4-year-old children were able to distinguish correctly between animals and machinery (Simons & Keil, 1995), they were not able to correctly assign plants to living objects until primary-school age (Berzonsky et al., 1988).

Learning environments in early childhood

Influential factors from learning environments in early childhood are categorised into structural and process features (Kluczniok et al., 2013; NICHD, 2002; Sylva et al., 2010; Tietze et al., 1998). Structural features represent the time-invariant and consistent conditions in the parental home or in kindergarten. Numerous studies have shown that features such as a lower SES, a lower level of parental education, or a migration background negatively affected young children’s language skills (Becker, 2011; Becker et al., 2013; Ebert et al., 2013; Sammons et al., 2009), their numeracy skills (Anders et al., 2012; Klibanoff et al., 2006; LeFevre et al., 2009; Melhuish et al., 2008; Sammons et al., 2009), or their SL (Hahn & Schöps, 2019; Martin et al., 2012; Morgan et al., 2016; Saçkes et al., 2011). However, the structural features of the kindergarten, such as spatial size, qualifications of the educational staff, group size and composition, or the number of toys, learning materials, or books, had positive effects on language or numeracy skills (Anders et al., 2012; Ebert et al., 2013; Kluczniok et al., 2013; Sylva et al., 2010).

The structural features of the parental home or kindergarten are related to the process features of these learning environments, which in turn affect the cognitive development of young children (Tietze et al., 1998). Process features include all learning-promoting activities and interactions with other children, parents, teachers, and the environment (Kluczniok et al., 2013; NICHD, 2002; Sylva et al., 2010; Tietze et al., 1998). On the one hand, previous findings have demonstrated that children with less beneficial structural features (e.g. low SES, having parents with lower levels of education, or having a different mother tongue) also had a lower quality of process features at home than children with better framework conditions (Bradley et al., 2001; NICHD, 2005; Sylva et al., 2010). In addition, previous findings have shown that the process features of the home learning environment (HLE) can directly contribute to higher language skills (Ebert et al., 2013; Melhuish et al., 2008; NICHD, 2005), higher numeracy skills (Anders et al., 2012; LeFevre et al., 2009; Melhuish et al., 2008; Niklas & Schneider, 2017), and SL (Morgan et al., 2016). These results indicate that especially children with less beneficial structural features at home are disadvantaged regarding their cognitive development through the lack of a high-quality HLE. On the other hand, studies have shown that these children benefit more from a higher quality of process features in early child care with regard to their cognitive, language, and numeracy skills than children with beneficial structures at home (Anders et al., 2012; Becker, 2011; Dearing et al., 2009; NICHD, 2002; Sylva et al., 2010). In addition to the HLE, early child care and its quality are also important for the cognitive development of young children and can even compensate for social and migration-related disparities.

Research questions

The aim of this study was to narrow the research gap concerning the development of the SL of kindergarten children in Germany and to examine whether there are differences in initial SL and, if so, whether these differences persist into primary school. Moreover, we aimed to determine possible starting points for a promotion of early SL by examining the structural and process features of the parental home and kindergarten. Therefore, we used latent growth curve models (LGCMs) to analyse longitudinal data from NEPS Starting Cohort 2 to answer the following research questions:

1. How does the SL of children develop from kindergarten to the third grade (LGCM 1)?

2. How do variables of the parental home and kindergarten affect initial SL and its development in children from kindergarten to the third grade (LGCM 2)?

In line with the results of Morgan et al. (2016) in the United States, we expected to find a (linear) growth in the SL of children from kindergarten to the third grade. Morgan et al. (2016) also showed that there were early disparities in SL, which persisted through primary school. We therefore expected to find similar results for this German sample.

In line with previous results from studies about numeracy, language, and scientific skills (e.g. Anders et al., 2012; Becker, 2011; Hahn & Schöps, 2019; LeFevre et al., 2009; Morgan et al., 2016; NICHD, 2002, 2003; Saçkes et al., 2011; Sammons et al., 2009), we expected to find significant effects of the structural and process features of the parental home and kindergarten on initial SL; in particular, we expected the social and linguistic background of the child to especially affect SL. In line with Morgan et al. (2016), we also expected to find significant effects on the growth of SL, supporting previous results that showed that some children who start with lower SL can partially compensate for this deficit.

Materials and methods

Sample and research design

To investigate the development of SL, we analysed data from the NEPS Starting Cohort 2 (Blossfeld et al., 2011), which included three measurement points (MPs): one in kindergarten (2011), one in Grade 1 of primary school (2013), and one in Grade 3 of primary school (2015). Starting with N = 2,937 children in kindergarten (49% female, aged 4.25–6.01 years), only N = 534 children were retested in Grade 1 of primary school (51% female, aged 6.17–7.75 years), and N = 479 were retested in Grade 3 of primary school (50% female, aged 7.92–9.42 years). Due to technical problems in the sampling procedure, which made it impossible to track a large proportion of the original sample between the first and second MP, N = 2,403 children were tested only at the first MP in kindergarten. From the remaining children, N = 450 participated in all of the three MPs. As a consequence, we decided to use multiple imputation (MI) to handle the missing data (see Statistical Analyses).

Measure

Scientific literacy

SL was measured using the NEPS tests for science in kindergarten, Grade 1, and Grade 3 (Hahn et al., 2013; Kähler, 2019a, 2019b; Schöps, 2013). All tests contained the same types of response formats: simple multiple choice and complex multiple choice in the special form of true-false items (see Figure 1). The NEPS science test for Grade 3 also included two items with a short constructed-response format. In all NEPS science tests, test administrators read out the questions and possible answers. In the NEPS science test for 4- to 6-year-old children (kindergarten) and the NEPS science test for children in Grade 1, the answers were picture-based, so children did not have to read or verbalise their answers, but had to show them (see Figure 1). For the NEPS science test in Grade 3, two items also contained texts as answers. All tests were carried out in individual testing sessions and had a time limit of 30 min (kindergarten and Grade 1) or 29 min (Grade 3).

Figure 1. Example of a NEPS multiple-choice item measuring scientific literacy in kindergarten (context: environment, component: development).

NEPS provides linked weighted maximum likelihood estimates (WLEs) for SL, which can be used in longitudinal analysis. These WLEs were estimated in a one-dimensional model based on the item response theory for all three MPs in ConQuest 4.2.5 (Adams et al., 2015). Because all three tests included different items, which were constructed to accurately measure SL within each age group, an anchor-group design was used to link the test scores from kindergarten and Grade 1 (Kähler, 2019a) and from Grade 1 and Grade 3 (Kähler, 2019b). This means that all items from two SL tests (kindergarten/Grade 1 and Grade 1/Grade 3) were administered in an independent link sample at a single MP. The responses to these items were used to link the respective tests and to estimate linked WLEs on a common scale. For further information about the linked tests, see Kähler (2019a, 2019b), and about the linking method, see Fischer et al. (2016). The NEPS science test for 4- to 6-year-old children contained 25 questions and had a WLE reliability of .75 (Schöps, 2013). The NEPS science test for Grade 1 also included 25 questions and had a WLE reliability of .73 (Kähler, 2019a). The NEPS science test for Grade 3 consisted of 22 items and had a WLE reliability of .68 (Kähler, 2019b).

Covariates

Structural features were divided into parental home and kindergarten. For the parental home, we considered the language spoken at home (1 = German, 2 = other language) and the migration background of the child (1 = no parent born abroad, 2 = one parent born abroad, 3 = both parents born abroad). We also included the number of siblings. The educational level of the parents was measured with the International Standard Classification of Education (ISCED; Schroedter et al., 2006), ranging from 1 = kindergarten to 6 = tertiary education. The SES of the family was measured with the International Socio-Economic Index of Occupational Status (ISEI; Ganzeboom et al., 1992), which determines the SES of the parents by their profession. For each feature (ISCED and ISEI), the highest specification was used. We also considered the monthly income (1 = less than €1,000–6 = more than €5,000). Additionally, we used the number of books at home (1 = 0 to 10 books to 6 = more than 500 books) as a proxy variable to measure the parental investment in cultural capital. We also included the structural features of the kindergarten as reported by the management of the kindergarten. We considered composite features of the group such as the proportion of children who spoke a language other than German at home, the proportion of children with a migration background, and the proportion of children with a low SES. Moreover, we included the educator-child ratio, the number of rooms, and the (scientific) focus of the kindergarten (1 = no scientific focus, 2 = scientific focus).

Process features from the parental home and kindergarten were also taken into account. This information was provided by the parents or the kindergarten educators of the child (accessible at https://www.neps-data.de/). For the parental home, we measured common activities (9 items, Cronbach’s α = .67) and learning opportunities (5 items, Cronbach’s α = .63) at home with an 8-point Likert scale ranging from 1= never to 8 = several times a day. For the kindergarten, we included common activities (10 items, 1 = never to 8 = several times a day, Cronbach’s α = .77), visits to external locations (8 items, 1 = never to 6 = nearly every day, Cronbach’s α = .67), and materials and toys (14 items, 1 = unavailable to 4 = available for almost all children, Cronbach’s α = .83). However, these items mainly included numeracy- and language-promoting activities and did not include many scientific activities. Means were computed for these five scales, which were then used in the MI and other analyses. Additionally, we included gender as a control variable in the analyses and also considered the time spent in kindergarten (years x hours per week; unit in years).

Statistical analyses

To handle the missing data, we applied MI. MI enables valid parameter estimation for missing data that are missing at random (MAR; Grund et al., 2016), whereas listwise deletion only offers unbiased parameter estimation when the missing data are completely at random. Thus, MI is less restrictive. Moreover, MI makes it possible to make use of all the available data, while taking the uncertainty caused by missing data into account. MI also provides the opportunity to take the multilevel structure of the data into consideration (data from the parental home as Level 1 and kindergarten as Level 2; Grund et al., 2016). Therefore, we ran the MI (N = 100 data sets, 50 iterations) in RStudio V3.5.2 (RStudio Team, 2015) with the covariates from the parental home and kindergarten, some auxiliary variables, the controls, and the competence scores for SL from all three MPs while considering the multilevel structure of the data.

To analyse the development of SL, two linear LGCMs were estimated in Mplus 7.4 (Muthén & Muthén, 2012). In these models, the WLEs were used to estimate the intercept and slope of SL from the three MPs. The loadings of the intercept were fixed to one. The slope was estimated as a linear growth fixed in kindergarten to zero, in Grade 1 to one, and in Grade 3 to two. The correlation between intercept and slope was allowed. Within the LGCMs we also took the given weights for the target into account. Moreover, we used the TYPE = COMPLEX option in Mplus to correct the standard errors. In LGCM 1, only the intercept and slope of SL were examined. LGCM 2 additionally included the covariates from the parental home and kindergarten on the intercept and slope of SL.

Results

Descriptives

Table 1 shows the descriptive statistics of all variables used in the LGCMs. The descriptives for the structural and process features of the parental home and kindergarten represent the scores from the first MP. The dependent variable SL has three scores for the three MPs (kindergarten, primary school Grade 1, and Grade 3). Moreover, Table 1 shows the missing rates of all variables, indicating the problem in the sampling procedure mentioned above. About 82% of the sample was missing at the second MP, and 84% was missing at the third MP.

Table 1. Descriptive statistics.

Variables	N	Min	Max	M	SD	Missings
Controls
Gender (1 = male, 2 = female)	2,937	1.00	2.00	1.49	0.50	0.00
Parental home
Language (1 = German, 2 = Other)	2,934	1.00	2.00	1.16	0.37	0.00
Migration background	2,296	1.00	3.00	1.43	0.73	0.22
Number of siblings	2,302	0.00	10.00	1.19	1.05	0.22
Parental education (HISCED)	2,511	1.00	6.00	4.26	1.41	0.15
Parental occupation (HISEI)	2,472	16.00	90.00	53.10	16.49	0.16
Number of books	2,438	1.00	6.00	3.73	1.32	0.17
Income (per month)	1,981	1.00	6.00	3.63	1.32	0.33
Activities	2,302	2.67	8.00	6.28	0.83	0.22
Learning opportunities	2,302	2.40	8.00	5.82	0.95	0.22
Time in kindergarten	2,394	0.01	1.81	0.57	0.28	0.18
Kindergarten
Children with another language (%)	2,446	0.00	95.00	23.57	25.12	0.17
Children with migration background (%)	2,049	0.00	93.00	28.57	22.25	0.30
Children with low SES (%)	1,886	0.00	80.00	20.15	20.25	0.36
Educator-child ratio	2,444	1.91	30.00	7.18	2.85	0.17
Number of rooms	2,511	2.00	73.00	13.31	7.63	0.15
Scientific focus (1 = no, 2 = yes)	2,568	1.00	2.00	1.32	0.47	0.13
Visits to external locations	2,734	1.00	3.50	2.25	0.42	0.07
Materials and toys	2,743	0.69	2.79	1.61	0.34	0.07
Activities	2,743	5.20	8.00	7.08	0.49	0.07
Competences
Scientific literacy kindergarten	2,937	−3.74	5.00	0.00	1.04	0.00
Scientific literacy Grade 1	534	−1.08	4.45	1.44	0.89	0.82
Scientific literacy Grade 3	479	0.37	6.10	2.80	0.93	0.84

Note: N = 2,937; SES = socioeconomic status.

LGCM 1: Scientific literacy

To examine the first research question, we used a linear LGCM to estimate the development of SL from kindergarten to Grade 3 of primary school. The fit of this model was good (RMSEA = 0.02, CFI = 1.00, SRMR = 0.02; Hu & Bentler, 1999). The significant mean of the slope of SL (M = 1.28, p < .001) showed that, on average, there was a linear increase in SL over time. Thus, the mean model-implied growth curve increased. Regarding the variance of the intercept, the significant result (s² = 0.68, p < .001) indicated that there was significant variability in the latent intercept of SL. Hence, children differed in their intercept in SL in kindergarten. However, the variance of the slope was not significant (s² = 0.07, p = .16), which means that there were no interindividual differences in the magnitude of the linear increase in the individual growth curves. Consequently, children did not differ in the extent to which their SL increased.

LGCM 2: Scientific literacy with covariates

To examine the second research question, which aimed to investigate the effects of the parental home and kindergarten variables on the intercept and slope of SL, we added these variables to the LGCM (RMSEA = 0.04, CFI = 0.91, SRMR = 0.02). Table 2 shows the results of the LGCM 2 (standardised effects). The size of the effects indicates the number of standard deviation units by which the SL changed when the independent variable changed by one standard deviation, after all of the other variables had been controlled for.

Table 2. Latent growth curve model 2: scientific literacy with covariates.

	Model 2
	Intercept SL		Slope SL
Controls	β	SE	β	SE
Gender (1 = male, 2 = female)	0.08**	0.03	−0.08	0.41
Parental Home
Language (1 = German, 2 = Other)	−0.24***	0.04	0.38	2.11
Migration background	−0.05	0.04	0.14	0.38
Number of siblings	−0.06	0.04	0.11	0.79
Parental education (HISCED)	0.14**	0.05	0.05	0.46
Parental occupation (HISEI)	0.04	0.05	0.03	0.41
Number of books	0.13**	0.04	0.17	0.56
Income (per month)	0.03	0.04	−0.06	0.59
Activities	0.01	0.04	−0.24	1.48
Learning opportunities	−0.03	0.04	0.04	0.37
Time in kindergarten	0.06	0.03	−0.08	0.65
Kindergarten
Children with another language (%)	−0.08	0.09	−0.22	2.33
Children with migration background (%)	0.00	0.08	0.28	2.11
Children with low SES (%)	−0.07	0.06	0.03	0.54
Educator-child ratio	−0.03	0.04	0.12	0.77
Number of rooms	−0.01	0.04	0.13	0.78
Scientific focus (1 = no, 2 = yes)	0.09*	0.04	−0.20	1.00
Visits to external locations	0.04	0.05	−0.02	0.25
Materials and toys	−0.01	0.04	0.03	0.31
Activities	0.00	0.04	0.10	0.27
R^2	.32		.46

Note: N = 2,937; SL = scientific literacy, β = standardised coefficients, SE = standard error.

*p < .05, **p < .01, ***p < .001.

Regarding the structural features of the parental home, there was a significantly negative effect of the language spoken at home (β = −0.24, p > 0.001) on the intercept of SL. A language other than German being spoken at home led to lower SL in kindergarten. Parents’ education had a positive effect (β = 0.14, p = 0.002), indicating that children whose parents had a higher educational level had higher SL. The same was true for the number of books (β = 0.13, p = 0.002). Children with a higher number of books in the parental home had higher SL in kindergarten than children with fewer books. Among the structural features in kindergarten that we examined, only the scientific focus had a significant effect on SL in kindergarten (β = 0.09, p = 0.02), indicating that children who attended a kindergarten that had a scientific focus had higher SL than children who attended a kindergarten that did not have a scientific focus. The process features, either from the parental home or from the kindergarten, did not have any significant effects on initial SL. Furthermore, neither the structural nor the process features had any significant effects on the slope of SL.

Effect sizes

In addition to LGCM 1, we used means and student-level standard deviations from the imputed data to compute standardised effect sizes to quantify the growth trajectories of SL (Hill et al., 2008). Examining the SL of all participating children, the effect sizes from kindergarten to Grade 1 (d = 1.23, CI [1.18, 1.29]) and from Grade 1 to Grade 3 (d = 1.25, CI [1.19, 1.31]) showed a strong and stable gain in SL. Children gained about 1.23 standard deviations on the SL test from kindergarten to Grade 1, and about 1.25 standard deviations from Grade 1 to Grade 3.

On the basis of the results from the second LGCM, we also estimated standardised effect sizes for the growth of SL separately for the significant variables language-use, parental education, number of books at home, and scientific focus in kindergarten. Figures 2, 3, 4 and 5 display the SL growth trajectories (means of WLEs in logits) for each group. In order to identify whether individual groups of children varied in their SL growth, we examined the means and estimated effect sizes.

First, we examined children who spoke a language other than German at home compared to children who spoke German at home (Figure 2). On average, children who spoke German at home had higher SL scores at all three MPs. However, children who spoke a language other than German at home had slightly reduced this gap between kindergarten and Grade 1. The effect sizes also showed this trend: the effect size for children who spoke German at home from kindergarten to Grade 1 (d = 1.13, CI [1.07, 1.19]) was clearly lower than the effect size for children who spoke a language other than German at home (d = 1.77, CI [1.62, 1.92]). From Grade 1 to Grade 3 the effect sizes were more similar (German: d = 1.32, CI [1.26, 1.38]; another language: d = 1.36, CI [1.22, 1.50]), indicating approximately parallel growth trajectories.

Figure 2. Development of scientific literacy: means of children who spoke German at home and children who spoke a language other than German at home.

Moreover, we looked at the means of children whose parents had different educational levels (Figure 3), which was measured with the highest ISCED. The ISCED has six levels of education, ranging from primary education to tertiary education. Children whose parents had tertiary-level education had, on average, higher SL scores at all MPs. This was also the case for children in kindergarten and Grade 1 whose parents had primary- or secondary-level education. In Grade 3, these groups were able to reduce the gap and therefore had more similar SL scores.

Figure 3. Development of scientific literacy: means of children whose parents had different educational levels (HISCED).

The effect sizes for children whose parents had different educational levels can be found in Table 3. On average, children whose parents had lower educational levels showed higher effect sizes and, thus, a higher growth in their SL. But, they were not able to fully close the initial gap between their SL and that of children whose parents had tertiary-level education.

Table 3. Effect sizes for the growth trajectories of SL: children whose parents had different educational levels (HISCED).

	n	MP	d	Lower CI	Upper CI
Primary education	39	Kiga to Grade 1	1.49	0.99	1.99
		Grade 1 to Grade 3	1.78	1.26	2.31
Secondary education I	203	Kiga to Grade 1	1.47	1.25	1.69
		Grade 1 to Grade 3	1.87	1.63	2.10
Secondary education II	931	Kiga to Grade 1	1.16	1.06	1.26
		Grade 1 to Grade 3	1.45	1.35	1.55
Post-secondary education	412	Kiga to Grade 1	1.16	1.02	1.31
		Grade 1 to Grade 3	1.16	1.01	1.31
Tertiary education I	490	Kiga to Grade 1	1.24	1.10	1.38
		Grade 1 to Grade 3	1.36	1.22	1.50
Tertiary education II	862	Kiga to Grade 1	1.29	1.19	1.40
		Grade 1 to Grade 3	1.22	1.12	1.32

Note: N = 2,937; MP = measurement point, d = effect size.

We also estimated the effect sizes of the growth trajectories of the SL of children with different numbers of books at home. The number of books served as a proxy variable to measure the parental investment in cultural capital. Categories ranged from 0–10 books to more than 500 books. Figure 4 shows the SL means for the six categories, and Table 4 shows the corresponding effect sizes. On average, children with more books at home had higher SL scores at the three MPs, although some groups overlapped at the second and third MPs. Especially from Grade 1 to Grade 3, children with 101–200 books had comparable SL means to children with more books. The effect sizes also showed that growth trajectories were higher from Grade 1 to Grade 3 than from kindergarten to Grade 1 (except for children with 201–500 books).

Figure 4. Development of scientific literacy: means of children with different numbers of books at home.

Table 4. Effect sizes for the growth trajectories of SL: children with different numbers of books at home

	n	MP	d	Lower CI	Upper CI
0–10 books	131	Kiga to Grade 1	1.37	1.10	1.64
		Grade 1 to Grade 3	1.75	1.46	2.03
11–25 books	310	Kiga to Grade 1	0.97	0.80	1.13
		Grade 1 to Grade 3	1.65	1.47	1.83
26–100 books	978	Kiga to Grade 1	1.27	1.17	1.37
		Grade 1 to Grade 3	1.35	1.25	1.45
101–200 books	596	Kiga to Grade 1	1.19	1.06	1.31
		Grade 1 to Grade 3	1.51	1.39	1.64
201–500 books	612	Kiga to Grade 1	1.34	1.21	1.46
		Grade 1 to Grade 3	1.16	1.03	1.28
More than 500 books	310	Kiga to Grade 1	1.27	1.10	1.44
		Grade 1 to Grade 3	1.19	1.02	1.36

Note: N = 2,937; MP = measurement point, d = effect size.

Finally, we estimated the effect sizes for children who attended a kindergarten that had a scientific focus and for children who attended a kindergarten that did not have a scientific focus. Figure 5 shows the SL means for both groups. On average, children in kindergartens that had a scientific focus had higher SL scores in kindergarten and Grade 1. In Grade 3, both groups had very similar average SL scores. This can also be seen in the effect sizes. Whereas the effect sizes from kindergarten to Grade 1 were similar (children in kindergartens that did not have a scientific focus: d = 1.19, CI [1.13, 1.26]; children in kindergartens that had a scientific focus: d = 1.17, CI [1.07, 1.27]), children in kindergartens that did not have a scientific focus showed a slightly higher effect size (d = 1.35, CI [1.28, 1.42]) and, thus, a higher growth trajectory from Grade 1 to Grade 3 than children in kindergartens that did have a scientific focus (d = 1.24, CI [1.15, 1.34]).

Figure 5. Development of scientific literacy: means of children who attended kindergartens with and without a scientific focus.

Discussion

The aim of this study was to investigate the SL of kindergarten children in Germany and its growth up to the third grade in primary school. We analysed whether there are early disparities in children’s SL and whether these initial differences persist into primary school. Moreover, we provided insights into the effects of structural and process features on SL and its growth. Such insights could indicate possible starting points for the promotion of SL, which could reduce or compensate for early disparities in kindergarten.

For the first research question, we examined how SL develops from kindergarten to the third grade of primary school (LGCM 1). The results confirmed the linear growth of SL. Thus, children increased their SL from kindergarten to the third grade. The effect sizes found in our analyses supported this result. They also showed strong and stable growth trajectories between the MPs. The LGCM also showed differences in the SL of children in kindergarten, which confirms previous results (Hahn & Schöps, 2019; Morgan et al., 2016). This means that there were already differences between the SL of kindergarten children before they entered school. To our surprise, there were no interindividual differences in the growth of SL. This means that the initial differences found in kindergarten persisted into primary school but neither decreased nor increased. Children with lower SL did not show more SL growth, and therefore, were not able to compensate for the initial differences. Attending primary school obviously was not enough to support children with early disparities and to help them to close the gap between them and their peers with higher SL. Our results are comparable with those of Morgan et al. (2016) and illustrate the importance of early promotion und support, especially for children with lower SL. To get the best support for young children, this promotion should already start in kindergarten and should be followed up in school.

In this context, the main limitation of this study should be mentioned. Due to the problematic sampling procedure, only 16% of the children were tested at all three MPs; 82% were tested only at the first MP. This problem in the sampling procedure occurred because the kindergarten management was asked to name the schools the children would attend after they finished kindergarten. Unfortunately, most children went to other schools than had been expected. It was not possible to track these children; this is the reason for the large amount of missing data. The exclusion of these cases would have led to biased parameter estimation. Because the data were missing due to this problem (at all kindergartens), we assumed that these cases were at least MAR, if not missing completely at random (MCAR). Because of the MAR assumption, we used MI (Grund et al., 2016). To evaluate the quality of the MI, we checked the convergence of the imputation model in which we used a variety of background variables from the first MP (from Level 1 and Level 2), and we also included some auxiliary variables in the imputation model (Grund et al., 2016; Zinn & Gnambs, 2018). Nonetheless, a distortion of the effects cannot be completely ruled out. This means that the results for the slopes should be interpreted carefully, and further longitudinal studies are needed to verify our results. Nevertheless, the results of this study can be seen as a first indication of persisting disparities in SL and they support previous results (Morgan et al., 2016).

For the second research question, we examined the effects of structural and process features on SL and its growth (LGCM 2) to find possibilities for reducing or compensating for the early differences in the SL of kindergarten children. Due to the low variance of the slope, the examined features did not have any significant effects on the growth of SL. However, there were significant effects on the initial SL in kindergarten. The language spoken at home had a negative effect, which is in line with results from studies on numeracy and language skills (e.g. Anders et al., 2012; Ebert et al., 2013). Children who spoke a language other than German at home had lower SL in kindergarten. This is also shown in the effect sizes. Although children who spoke a language other than German at home showed stronger growth in SL from kindergarten to Grade 1, they were not able to close the initial gap between their SL and that of children who spoke German at home. These results show the importance of the linguistic background of the child for SL (Hahn & Schöps, 2019; Morgan et al., 2016) and indicate a possible starting point for improving young children’s acquisition and development of SL. Supporting children who speak a language other than German at home as early as possible may help to reduce these initial disparities in SL and could offer them the same access to learning in school as those of children who speak German at home. Further longitudinal studies should clarify whether the early and systematic support of these children, particularly children from socially disadvantaged homes, can counteract the early effects and manifestations of social and migration-related disparities and also contribute to an improvement in SL.

In addition to the language spoken at home, the parental education also had a significant effect on the initial SL. Children whose parents had a higher educational level also had higher SL. This finding corresponds to results from various studies on numeracy and language skills (Anders et al., 2012; Becker et al., 2013; LeFevre et al., 2009; Melhuish et al., 2008; NICHD, 2002). In particular, children whose parents had tertiary-level education had higher SL at all MPs. Moreover, the effect sizes found in our study revealed that children whose parents had post-secondary-level education or lower had higher growth trajectories, in particular from Grade 1 to Grade 3. Thus, these children were able to reduce the gap over time. Parental education seems to affect young children’s SL, especially in kindergarten and in Grade 1 of primary school. It could be assumed that, in general, parents with a higher educational level also give education a higher priority and, therefore, promote and support their children’s development and their acquisition of SL.

That is why the number of books at home (as a proxy measurement of the parental investment in cultural capital) is also a significant factor for young children’s SL. A positive effect of the number of books on the SL of young children has already been found for children in primary school (Martin et al., 2012). In our study, we were able to replicate this effect for children in kindergarten. The effect sizes for the growth trajectories of the number of books are comparable with those for parental education. Whereas children with a high number of books at home clearly had higher SL at all MPs, children with fewer books at home were able to slightly reduce the existing gap in kindergarten. However, on average, they still showed lower SL in Grade 3 than children with a high number of books at home. Altogether, these results demonstrate that the initial differences in SL can be attributed to social and linguistic disparities in early childhood.

Looking at the structural features of the kindergarten, a scientific focus had a significantly positive effect on SL. Attending a kindergarten that explicitly offered a scientific focus directly contributed to a higher level of SL. An example of kindergartens that have a scientific focus are ‘Little Scientists’ House’ kindergartens (Haus der kleinen Forscher; Anders et al., 2018). Kindergartens that participate in this German educational programme have a specific concept and have equipment specifically recommended for supporting young children’s SL. For example, educators can learn how to impart knowledge about science and research to young children and, thus, young children’s SL can improve.

Therefore, children who attend a kindergarten that has a scientific focus have more educational opportunities related to science in early childhood. A kindergarten that has a scientific focus can form the starting point from which children can go on to expand their formal scientific knowledge in school, from which they – in turn – go on to become highly qualified individuals in the field of science.

However, with regard to the effect sizes for the growth trajectories of children in kindergartens with and without a scientific focus, these initial differences seemed to disappear in Grade 3 of primary school. Children in kindergartens that did not have a scientific focus were able to close the gap between them and their peers who attended a kindergarten that did have a scientific focus. Thus, a scientific focus in kindergarten seemed to have only a short-term effect on SL. We assume that the structured and systematic science education that started in primary school helped to reduce these early differences and brought the children from different kindergartens onto the same level of SL. In Germany, there is no specific science curriculum in kindergarten, and children learn very differently about science. That is why the first challenge for schools (regarding children’s SL) is to bring all children onto the same level of SL to create equal conditions for science teaching in secondary schools. In order for science teaching in secondary schools to be successful, the basis for scientific thinking and scientific knowledge must already be formed in primary school (Möller et al., 2002). Building on these results, it would be interesting to examine the influence of different ways of teaching and specific activities at school on the development of children’s SL. Unfortunately, this was beyond the scope of our study but it is an important task for future research.

Nevertheless, attending a kindergarten that has a scientific focus can bring children closer to the natural sciences and can thus create a basis for the development of SL as early as possible. However, further research is needed to examine these specific effects. In particular, further research should investigate the quantity and quality of specific scientific contents or activities in kindergartens that have a scientific focus. This could help to identify the relevant process features in kindergartens that specifically affect young children’s SL, so that other kindergartens that do not have a scientific focus could benefit from these findings and implement comparable activities. Our study revealed that the activities that measured the quality of the parental home and kindergarten did not have a significant effect on the SL of kindergarten children. We assume that these activities (mainly numeracy- and language-promoting activities) did not address enough scientific activities, which can be seen as a limitation of this study. Moreover, the examination of more science-related process features in kindergartens could help to reveal the meaning of the label ‘scientific focus’. Due to the limited NEPS data, the binary coded scientific focus was the only science-related variable. The management of the kindergarten could choose between being a kindergarten with a scientific focus or not. One can assume that all educators explore scientific concepts in their daily routine in kindergarten, regardless of whether they work in a kindergarten that has a scientific focus or not. The questions of how the educational behaviour of educators at kindergartens that have a scientific focus differs from that of educators at kindergartens that do not have a scientific focus remains open, as does the question of how that difference affects young children’s SL. Future studies should help to answer these questions.

Conclusion

The results of this study show that young children already differed in their SL in kindergarten and that these differences were based on social and linguistic disparities. Moreover, our results reveal that the SL of these children grew up to the third grade of primary school but that there were no differences in the growth. This means that children from socially disadvantaged homes or children who spoke a language other than German at home had lower levels of SL and were not able to compensate for these deficits. These disparities were still found in the third grade of primary school, which revealed that attending primary school did not in itself enable the children to reduce these initial deficits. Promoting the acquisition and development of SL as early as possible could help to reduce these disparities and to prepare these children for their later life.

Availability of data and materials

The data sets generated and analysed in the current study are available in the NEPS repository: https://www.neps-data.de/

Disclosure statement

No potential conflict of interest was reported by the author(s).