ToM is the awareness and reasoning of others' and one's own mental states (for a comprehensive review, see Hughes and Devine, 2015). Associated with cognitive correlates such as executive function and language, as well as social correlates such as family, peers, and cultural contexts, ToM development is evident from infancy to adolescence and most research shows that around the age of 4, children realize that people's behavior is guided by their beliefs that might be different from their own beliefs or the reality. Inherently related to knowledge formation, understanding how the mind works is closely associated with children's interpretation of the nature of teaching and learning (Olson and Bruner, 1996; Kuhn, 2000; Wellman and Lagattuta, 2004). Unfortunately, ToM research in developmental psychology predominately focuses on the acquisition of false belief understanding during early childhood (Wellman et al., 2001). As a result, our understanding of the relationship between ToM and children's comprehension of teaching and learning is limited.

Rooted in the inquiry of the nature of knowing and knowledge, personal epistemology explores people's thoughts on what knowledge is and how knowledge is constructed and evaluated (Burr and Hofer, 2002; Kuhn and Weinstock, 2002). Kuhn (1999) distinguished the levels of epistemological thinking before and after children achieve the false belief understanding as the realist level and the absolutist level. The realist level of epistemological thinking perceives people's assertions as copies of an external reality, whereas the absolutist level sees the assertions as either correct or incorrect representations of the reality. The transition from the realist level to the absolutist level is a decisive moment in the development of epistemological thinking. It indicates the “transition from simple, unconscious, unreflective knowing about the world to a second-order, or metacognitive, reflection on the knowing claims of self and others” (Kuhn and Weinstock, 2002, p. 126).

Understanding the mental process of teaching and learning is vital for becoming metacognitive self-regulated learners. By working definition, learning, and teaching in the current context refer to acquisition of explicit knowledge instead of tacit knowledge. Adopting a mentalisitic perspective, Ziv and Frye (2004) defined teaching as “an intentional activity to increase the knowledge (or understanding) of another, thereby reducing the knowledge difference between teacher and learner” (p. 458). Learning, on the other hand, is not necessarily intentional, although it entails knowledge change similar to that in teaching (Wang, 2010).

There is evidence that children's understanding of teaching and learning develops during preschool years, alongside with their emerging ToM ability (Ziv and Frye, 2004; Frye and Ziv, 2005; Ziv et al., 2008; Wang, 2010). Ziv and Frye (2004) developed stories of different teaching scenarios to probe children's understanding of the knowledge state in teaching. They found 3-year-old children responded that the ignorant person should be taught, and the knowledgeable person should teach, regardless of their other characteristics such as age or occupation status. However, when the teacher overestimated his or her own knowledge or the learner's knowledge, only 5-year-old children who had passed the false belief task predicted that the teacher's beliefs, but not the actual knowledge state, would dictate the teaching behavior. Children's performance on the teaching tasks involving teacher's false beliefs correlated with their performance on the false belief task, r = 0.31, controlling for age. This result was since replicated in French children, with a correlation between teaching stories involving false belief and ToM at 0.23 (Bensalah et al., 2012).

Children's performance on false belief tasks was also found to be related with their understanding of teaching intentions. In another series of studies (Frye and Ziv, 2005; Ziv et al., 2008), children were told stories either involving a teaching intention or not, and asked to judge whether the teacher intended to teach. For example, in the successful imitation story, the teacher was unaware of the presence of the learner, hence no teaching intention; in the embedded teaching story, the teacher disguised her teaching as playing a game instead of explicitly labeling it as an intentional teaching event. Again, only 5-year-old children could distinguish the teaching intention from the learning intention in the imitation story, or detecting a teaching intention embedded in a game. The understanding of teaching intentions was correlated with children's false belief understanding after the effect of age was partialed out, with r ranging from 0.33 to 0.59 across various tasks.

Focusing on the concept of learning, Wang (2010) developed a set of stories to examine the relation between ToM development and children's understanding of knowledge change and intention in learning. It was demonstrated that children's understanding of the learning concept underwent a transformation from a behaviorist perspective to a mentalistic perspective from 4- to 6-years of age. In the coincidence story for example, one protagonist was told that the circle she drew was exactly the letter O, while the other protagonist was simply complemented what a good job he did in drawing the circle. Children were asked to judge which one of the two characters learned how to write a letter O in the story. The results showed that young children responded in random, indicating a behaviorist view of learning. Similar to teaching comprehension, they also failed to appreciate the effect of false belief on learning. When the learner overestimated or underestimated own knowledge, younger children found it difficult to decide whether the learner needed to learn, or whether the learner would try to learn. Younger children also encountered difficulty in understanding learning stories with a conflict between the learning intention and the learning outcome. They over-attributed learning intentions to learners who learned something either by accident (hence a discovery learning) or through implicit learning without an initial intention. Changes in children's understanding of learning were correlated with their emerging ToM ability, with partial correlations ranging from 0.24 to 0.36 controlling for age across different tasks.

The small to medium effect sizes of the association between ToM and teaching and learning comprehension reviewed above suggested that these are two distinctive yet closely related constructs. The relation could be bidirectional: on the one hand, advanced mental state reasoning might help children to process teacher's intention and reflect on their learning; on the other hand, children's conceptualization of teaching and learning from everyday experience, especially that in formal schooling context, might provide children with more opportunities to discuss their mental state and practice mental state reasoning, hence lead to more advanced ToM.

This latter view is supported by the social perspective of ToM (Hughes and Devine, 2015) that emphasizes the environmental influence on mental state reasoning development. In the last two decades, various lines of inquiry have demonstrated that social input enriched with mental state discourse advances children's ToM development. Such inquiries include mental state discourse training in classrooms (e.g., Lecce et al., 2014), conversational environment of deaf children (e.g., Peterson and Siegal, 1995), and family characteristics including maternal sensitivity (e.g., Meins et al., 2003), content, and quality of family conversation (e.g., Jenkins et al., 2003), and sibling relationship (e.g., Hughes et al., 2010). In the case of schooling experience, Flavell (1987) claimed 30 years ago that the epistemic discourse in teaching and learning context could be a “hotbed” (p. 27) for the acquisition of false belief understanding. Surprisingly, little empirical evidence exists to date demonstrating how and to what extent the formal schooling affects children's ToM development, with the exception of some indirect indications. For example, it is suggested that early access to formal schooling might lead to more advanced mental state understanding (e.g., Bensalah et al., 2012). Indeed, cross-cultural comparison found British children who started formal schooling at an early age outperformed Italian and Japanese children on false belief understanding (Hughes et al., 2014). A more recent cross-cultural study demonstrated that comparing to British children, Hong Kong children attending local primary schools which predominately adopted the drill-and-practice pedagogy demonstrated delay in ToM use, while children attending international schools in Hong Kong adopting an inquiry based pedagogy were not delayed (Wang et al., 2016), even though in both local schools and international schools Hong Kong children outperformed their British counterparts in executive function, a cognitive stipulation for ToM (Devine and Hughes, 2014). The finding hinted the quality of epistemic discourse in schools plays an important role in children's ToM use in middle childhood.

Therefore, the question boils down to whether early comprehension of teaching and learning affects children's ToM development, and to what extent. The primary goal of the current study is to empirically test the relationship between children's teaching and learning comprehension and their ToM development by consolidating and validating a comprehensive teaching and learning scale for young children based on previous studies. More importantly, the current study goes beyond previous research and conducts causality analysis based on empirical data to answer the direction of causality question in the relation between teaching and learning comprehension and children's ToM ability. In order to achieve this goal, however, it is necessary to consider the cultural context of both ToM development and teaching and learning comprehension. Modern schooling as a rather recent phenomenon in our evolutional history functions as an important, if not the most important, institution for cultural transmission (Bjorklund and Bering, 2002). Schooling experience as a form of social input is undeniably culturally specific (Gutiérrez and Rogoff, 2003). Hence our conceptualization of teaching and learning is a product of both cultural beliefs and schooling practice.

Both ToM and teaching and learning comprehension are arguably developed in cultural context. Even though children from different countries develop false belief understanding between 3- and 5-years of age (Wellman et al., 2001), there are remarkable cultural differences in the rate and trajectory of ToM development during early childhood (e.g., Naito and Koyama, 2006; Wellman et al., 2006; Liu et al., 2008; Shahaeian et al., 2011; Hughes et al., 2017). Both Japanese children (Naito and Koyama, 2006) and Hong Kong children (Liu et al., 2008; Hughes et al., 2017) were reported to show delayed false belief understanding acquisition comparing to Western children. Chinese (Wellman et al., 2006) and Iranian (Shahaeian et al., 2011) children develop knowledge understanding prior to belief understanding measured by a ToM scale, contrary to that of North American and Australian children.

Conceptualization of teaching and learning could be culture specific depending on cultural belief and context. Explicit knowledge transmission especially in the formal schooling system was not always the norm. Rogoff et al. (1993) reported that Mayan children rely exclusively on observing more experienced adults to learn. In such an observational learning context, it is probably justifiable to conceptualize learning as doing instead of knowing. In another light, while learning means acquiring knowledge and understanding in Kantian philosophy, Li (2001, 2002) argues that learning in Confucius philosophy implies virtue, or perfecting oneself, in addition to its epistemic nature. Supporting this view, Wang (2010) found that Chinese children insisted on the person who already learned certain knowledge should keep on “learning” the same knowledge again and again for the purpose of practice and achieving perfection.

In addition to the virtuous nature of learning, Asian cultures, especially Chinese cultures, highly value persistence and effort in children's learning. Asian children spend longer time in schools (Stevenson et al., 1990), and longer hours doing homework (Chen and Stevenson, 1989). Comparing to that in the West where formal academic learning usually starts around the age of five and six or even later, early childhood education in Hong Kong is heavily influenced by formal education practice (Pearson and Rao, 2006). Hong Kong children start formal numeracy and literacy learning as early as 3 years of age in kindergartens, often with daily homework (Li and Rao, 2000, 2005; Ng, 2014; Ho et al., 2015). As many of 81% of the kindergartens in Hong Kong teach two-digit addition, 42% of them teach two digit addition with carry-over (Cheng et al., 2001). A recent survey showed that 9-year-old Hong Kong primary school children on average spend 151 min every weekday doing homework (Ho, 2016). At home, Chinese parents are more involved in their children's learning, and demonstrate more formal teaching methods in their parenting (Pomerantz et al., 2014). Hong Kong government started to provide early childhood education vouchers to families since 2006, yet middle class parents tend to spend the extra cash to pay for extracurricular academic learning (Rao et al., 2016).

In sum, growing up in an intensely academic-oriented society, Hong Kong children are more likely to have a rather mature understanding of teaching and learning from an early age. This inspires another goal of the current study, which is to compare children's epistemic beliefs in the context of teaching and learning across cultures, in this case, Hong Kong and the United States. Going beyond the WEIRD (Western, Educated, Industrialized, Rich, and Democratic) societies that the majority of the research is based on, this is the first study of its kind examining cross-cultural differences in children's comprehension of teaching and learning, as well as the relation between teaching and learning comprehension and ToM development in a cultural context. The result would inform us the similarities and differences in children's personal epistemology across cultures, and highlight the role of cultural belief and practice as social input in ToM development.

In summary, we set out to meet three aims in this study. The first aim was to develop and validate a comprehensive Preschool Teaching and Learning Comprehension Index (PTLCI) based on tasks derived from previous studies investigating children's understanding of knowledge and intention in teaching and learning (Ziv and Frye, 2004; Frye and Ziv, 2005; Ziv et al., 2008; Wang, 2010), and establish measurement and structural equivalence of the PTLCI with a multilevel structural equation model (MSEM) (Muthén, 1994; Byrne and van de Vijver, 2014) using samples drawn from the U.S. and Hong Kong. The second aim was to examine the correlation between children's teaching and learning comprehension, measured with PTLCI, and their ToM development using a multiple-indicator-multiple-cause structural equation model (MIMIC-SEM) (Kline, 2016). In doing so it also enables us to examine the cross-cultural differences in the PTLCI and ToM development. Last but not the least, the third aim was to empirically evaluate the relationship between ToM and teaching and learning understanding and identify causal relationship through the MIMIC-SEM and reverse causality analysis.

Two groups of 4–6-year-old children participated in this study. The U.S. sample was consisted of 97 English speaking children including 42 girls and 55 boys (M = 64.6 months, SD = 9.63), recruited through kindergartens and schools serving middle class neighborhood in a large East Coast city of the United States. The majority of the sample was Caucasian. Two third of the mothers in the American sample (n = 61) had university level or above education. The Hong Kong sample was consisted of 106 Cantonese speaking children from Hong Kong including 58 girls and 48 boys (M = 63.8 months, SD = 9.81), recruited through kindergartens and schools serving middle class neighborhood in the New Territories of Hong Kong. All children in the Hong Kong sample were ethnic Chinese. Less than half of the Hong Kong mothers (n = 43) had university level or above education. The study was approved by the Human Research Ethics Committee of the Education University of Hong Kong and the Institutional Review Board of University at Buffalo, respectively. Written informed consents were obtained from parents and guardians before the data collection.

To validate a measurement universally adaptive and generalizable in both the and Hong Kong culture at the sample level as well as at the country level, we adopted the multilevel-multicultural perspective (Byrne and van de Vijver, 2014) which could address the disaggregated individual and aggregated cultural data to establish measurement equivalence and structural equivalence. Measurement equivalence refers to the equality of the mean of the manifest variable; while structural equivalence means the equality of the relationships among latent variables. By adopting a latent variable approach in contrast to the manifest variable approach (Selig et al., 2008), we aimed to decompose the total variance into within- and between-group components and enhance specifying separate structural model at each level (Byrne and van de Vijver, 2014).

Step 1 would be the construct validation of the PTLCI by confirmatory factor analysis (CFA) using item-parceling to reduce the number of parameters in order to improve model fit and decrease noises in investigating the structural models (Bandalos, 2002; Little et al., 2002).

Step 2 would establish a multilevel structural equation model (MSEM) (Muthén, 1994; Byrne and van de Vijver, 2014) to examine the measurement and structural equivalence across the country and the individual levels.

Step 3 and Step 4 would evaluate the relationship between ToM and teaching and learning understanding and identify causal relationship through reverse causality analysis. In Step 3, we adopted a multiple-indicator-multiple-cause structural equation model (MIMIC-SEM) (Kline, 2016). MIMIC-modeling is an integration of both formative and reflective modeling. In a formative model, observed variables are the causes of the latent variables. On the contrary, in a reflective model, latent variables are the causes of the observed variables. Hence, in a MIMIC-structural model, the independent observed variables serve as the causes of the latent variables which in turn contribute to the outcome dependent observed variables. Step 4 would target at ensuring causality of the structural models using reversed causality analysis. In a cross-sectional design, reversed causality analysis will confirm the directionalities of causations among the studied constructs by reversing the direction of causality and comparing the model fitness indexes to that of the original model. The model with significantly better model fitness would be accepted as the causality model (Antonakis et al., 2010).

The analyses were composed of two phases. First, in order to ensure the construct validity of the PTLCI, the relationships among the items and the latent factors were evaluated by confirmatory factor analyses (CFAs) by means of maximum likelihood (ML) estimation. The rationale for item parceling was based on the face validity of the 16 PTLCI (Marsh et al., 1999). This step would be conducive to investigating and ameliorating the potential measurement problems. The item parcels of the 16 items were illustrated by Table 1 above.

Prior to the CFAs, multivariate and univariate data screening had been conducted involving the investigating outliers, skewness, and kurtosis for the sake of the normality assumption of CFA, although with a sample size larger than 200, the problem of skewness and kurtosis was not a major concern (Tabachnick and Fidell, 2001). The skewnesses of all the items fell between −2 and 2 while the kurtoses of all the items fell between −7 and 7 (George and Mallery, 2010; Table 2), hence all of the data were retained.

The means, SDs and the intercorrelations of and among all the studied variables were tabulated by Table 3 below.

The CFA result indicated that the 4-parcel model of the PTLCI fit the data sufficiently: χ² (df = 2) = 1.046, p = 0.5926, CFI = 1.00, SRMR = 0.0196, GFI = 0.997, RMSEA = 0.000. The result demonstrated excellent construct validity of the PTLCI. The aforementioned CFA model is depicted by Figure 1 below.

The focus of Step 2 was to establish measurement and structural equivalence of the PTLCI between and within each country. The MSEM analytical strategies purported by Muthén (1994) and Byrne and van de Vijver (2014) were employed in this study. A two-level Structural Equation Model (SEM) was adopted to include the between country level and the within country (individual) level, taking age, and language ability into consideration. As mentioned above, in MSEM, the overall variance could be decomposed to the group and individual levels. The between-country/within-country structural model was depicted by Figure 2 below, with good model fit: χ² (df = 32) = 20.200, p = 0.9477, CFI = 1.00, SRMR = 0.0343, GFI = 0.986, RMSEA = 0.000. Therefore, the aforementioned multilevel model as well as the structural equivalence were duly validated across both the country and individual levels. It was implied that the PTLCI as well as the ToM measure were both culturally adaptive in both Hong Kong and the U.S. Furthermore, in both regions, a child's PTLCI score would have a significant and positive contribution to the child's ToM capacity, with a standardized effect size of 0.466.

The focus of Step 3 was 2-fold: one, to capitalize on the association between PTLCI and ToM found in Step 2 using a MIMIC-SEM to determine causality; and two, to examine cross-cultural differences of the two constructs of interests by entering country in the model. The MIMIC-SEM result was illustrated in Figure 3. The model indicated that the hypothetical model fit the data sufficiently: χ² (df = 17) = 23.931, p = 0.1213, CFI = 0.960, SRMR = 0.0515, GFI = 0.971, RMSEA = 0.0448. The standardized indirect effect of the independent variables were illustrated by Table 4 below. The MIMIC structural model indicated that the children in Hong Kong outperformed their U.S. counterparts in the PTLCI (β = −0.606, p < 0.001). Moreover, according to the Table 4, Hong Kong children also outperformed their U.S. counterparts in ToM, with standardized indirect effect = −0.299, p < 0.001.

In order to investigate the causalities of the models, reversed causality analyses were conducted in this study. The results were illustrated by Table 5 below.

Among the validated and reversed structural models, ΔCFI ≥ 0.01 which reflected that the validated and reversed models were not equivalent (Cheung and Rensvold, 2002). Under the comparison of CFI, GFI, SRMR, and RMSEA, the initially validated models demonstrated more desirable model-fit indices and hence was found to be more appropriate in terms of causality.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.