Maternal mental state language is associated with trajectories of Chinese immigrant children’s emotion situation knowledge

Abstract

We investigated the role of mothers’ references to mental states and behaviors and children’s emotion situation knowledge (ESK) in a prospective, cross-cultural context. European American mothers (n = 71) and Chinese immigrant mothers (n = 60) and their children participated in the study. Maternal references to mental states and behaviors were assessed at Time 1 when children were three years of age. ESK was assessed when children were 3, 3.5, and 4.5 years of age. Multi-group latent growth curve analyses were used to model children’s growth in ESK over time, as well as relations between mental state language and references to behaviors on children’s trajectories. Results indicated that maternal references to mental states were associated with concurrent levels of ESK for European American children, and change over time for the Chinese immigrant children. Maternal references to behaviors were negatively associated with concurrent ESK for both groups.

Keywords

Mental state language emotion understanding culture behaviors

Introduction

Children’s emotional understanding is associated with socialization practices (Denham, 1993; Dunn, Bretherton, & Munn, 1987; Dunn, Brown, & Beardsall, 1991; Eisenberg, Cumberland, & Spinrad, 1998). In particular, mother–child conversations appear to be an important context in which children learn about situations that may elicit feelings and how emotions influence behavior (for a review, see Harris, 1989). For example, mothers’ use of mental state language (MSL) (references to desires, emotions, and cognitions) significantly contributes to children’s understanding, representation, and reasoning about psychological states (for reviews, see Carpendale & Lewis, 2004; Symons, 2004). When children are exposed to high levels of MSL, they showed better social cognitive abilities, including theory of mind (Ruffman, Slade, & Crowe, 2002) and emotional understanding (Doan & Wang, 2010; Taumoepeau & Ruffman, 2006).

A limitation of past research is that most of these studies have examined the role of MSL concurrently and in one cultural group, namely European Americans (EAs), leaving a gap in the literature about its long-term contribution and replicability across cultures. Furthermore, the majority of the supportive evidence comes from research with children in the early stages of development (roughly around 24–36 months) when the acquisition and expression of mental state terms are just beginning to emerge (Bartsch & Wellman, 1995). Studies that examined the associations between maternal MSL and children’s social cognitive abilities in the later preschool years appear to be more inconclusive with some studies having failed to replicate the finding consistently (see La Bounty, Wellman, Olson, et al., 2008; Ontai & Thompson, 2002).

Another limitation of previous research is the lack of cross-cultural studies. Understanding the role of culture is important due to evidence showing cultural differences in mother–child conversations (Fivush & Wang, 2005; Wang, 2001; Wang & Fivush, 2005), as well as in attitudes and beliefs towards emotions (Markus & Kitayama, 1991; Potter, 1988). Finally, few studies have examined both references to behaviors and MSL simultaneously. The purpose of this study is to examine associations between mothers’ use of MSL and references to behaviors and trajectories of children’s developing emotion situation knowledge (ESK) in EA and Chinese immigrant (CI) families.

Mental State Language and Children’s Social Cognitive Development

Parents play a critical role in their children’s linguistic environment from early development (Bloom, 2000). Parents from across cultures socialize emotion through language as a way of teaching children affective lexicon and culturally appropriate ways of interpreting and understanding emotional states (Hofmann & Doan, 2018; Ochs & Schieffelin, 1989). Parents may model or label their own and others’ mental states, provide feedback on their children’s use of mental state words, or inquire about mental states in conversation (Clancy, 1999; Welch-Ross, Fasig, & Farrar, 1999). Conversational references to mental states can increase attention to internal states and their associations with the physical world (Dunn, Brown, Slomkowski, et al., 1991; Ornstein, Haden, & Hedrick, 2004; Wellman, Harris, Banerjee, et al., 1995). Furthermore, when there are differences in emotions between the self and others, parents’ explanations using mental state terms make it explicit that there are multiple perspectives and differing beliefs and causes for emotions, which can facilitate children’s theory of mind development and emotional understanding (Harris, 1989). Overall, MSL focuses on the interpretation and evaluation of an event or the motivations of an agent, and children’s acquisition of MSL occurs along with the construction of their social world (Dunn, Bretherton, & Munn, 1987; Rudek & Haden, 2005; Wood & Conway, 2006).

Several studies to date have linked parents’ discussions of mental states to young children’s emotional understanding (Dunn, Brown, & Beardsall, 1991; McQuaid, Bigelow, McLaughlin, et al., 2008; Tarullo, Youssef, Frenn, et al., 2016; Taumoepeau & Ruffman, 2006, 2008). Mothers’ references to desires at 15 months predicted children’s later usage of mental state terms, as well as children’s emotional understanding at 24 months (Taumoepeau & Ruffman, 2006). In addition, Denham, Zoller, & Couchod (1994) found that mothers’ use of emotion discourse predicts 47-month-old children’s advanced identification of facial expressions and emotion-eliciting situations. Mothers’ use of empathy-related language has also been found to be related to children’s understanding of situational antecedents of emotion, independent of children’s use of emotion language (Garner, Jones, Gaddy, et al., 1997).

At the same time, some conflicting results have also emerged. Associations between children’s theory of mind at 4.5 years of age and mothers’ use of MSL were not found (Ontai & Thompson, 2008), although significant relations between maternal elaborative discourse and children’s theory of mind were present. And although mothers’ references and causal explanatory language about emotions at 20 months predicted children’s concurrent emotion understanding abilities, this association failed to be confirmed longitudinally at 68 months (La Bounty, Wellman, Olson, et al., 2008). Similarly, maternal cognitive and feeling talk with either 2- or 4-year-olds predicted only the younger children’s mental state talk two years later, but not older children’s MSL (Jenkins, Turrell, Kogushi, et al., 2003).

In sum, although concurrent associations are usually found among toddlers, the findings are often mixed with preschool children. Past studies also suffer from methodological limitations, such as the lack of data beyond one or two timepoints and the use of different measures of social cognitive skills across time. Having a minimum of three timepoints with the same measure will make it possible to assess predictors of both current abilities and the growth of these abilities over time (Preacher, Wichman, MacCallum, et al., 2008). Furthermore, most prior research has failed to consider the role of culture in shaping these processes.

Culture, Mental State Language, and Emotion

In essence, MSL serves to express one’s opinions, feelings, thoughts, and desires and it is a means to understand another’s mental state. Because the concept of the self is thought to vary across cultures, there is a reason to believe that cultural values and belief systems may influence the frequency with which MSL is used. Specifically, one way in which cultures have been differentiated is on the continuum of independence and interdependence, a social orientation that refers to the degree to which individuals are encouraged to focus on themselves (i.e., independence) or on other people (i.e., interdependence) (Markus & Kitayama, 1991). Although both independence and interdependence co-exist in any one culture (Mascolo & Li, 2004), different cultures prize and value different attributes (such as individuality versus communality). In turn, these cultural preferences influence how the self is socialized and, hence, constructed. This cultural self-construal may play an important role in both the use of MSL and children’s emotional development. For example, in an interdependent culture, individuals’ emotions are often not highlighted, as mutual group consensus is more valued over personal expression of emotions (Markus & Kitayama, 1991).

In the context of mother–child conversations, Wang and her colleagues have found that EA mothers often make more comments on mental states and are more likely to refer to the child’s personal needs, preferences, judgments, and opinions as compared to their Chinese counterparts (Wang, Leichtman, & Davies, 2000). During memory sharing of emotional events, Chinese mothers are more likely to focus on behavioral norms and social expectations, whereas EA mothers are more likely to give causal explanations for feeling states (Fivush & Wang, 2005; Wang, 2001). In addition, Tardif & Wellman (2000) found that although Chinese children acquired desire terms earlier than their EA peers, they acquired cognitive terms later (also see Han, Leichtman, & Wang, 1998). Compared with EA preschoolers, Chinese children living in China, and CI children in the USA show less advanced ESK (Wang, 2003; Wang, Hutt, Kulkofsky, et al., 2006).

The minimization of emotions and mental states in Asian cultures may shape family narrative practices, which, in turn, influence developmental outcomes pertaining to emotional understanding. However, to the best of our knowledge, no study has examined in a cross-cultural context the association among maternal references to mental states and behaviors and children’s trajectories of children’s emotion knowledge over time. In this study, we investigate children’s developing emotion knowledge across the preschool years and their association with mother’s use of MSL.

Methods

Participants

Seventy-one EA mother–child dyads (37 boys, 34 girls) and 60 CI dyads (30 boys, 30 girls) from a university town and suburban areas in upstate New York participated in the study. Data collection took place at three timepoints when children were on average 36 months (Time 1), 42 months (Time 2), and 54 months of age (Time 3). The mean age was not significantly different between the two cultures (see Table 1). Children were recruited from local schools and by word of mouth and were taking part in a larger longitudinal study of socio-cognitive development across the preschool years. All children came from middle-class backgrounds, with the majority of the mothers (CI, 98%; EA, 93%) having obtained a college degree or beyond. As a measure of socioeconomic status, maternal education was coded on a 3-point scale: 1 = high school qualification, 2 = four-year college degree, 3 = postgraduate degree. CI families were originally from mainland China, Hong Kong, and Taiwan, with the majority from mainland China (93%). Most (80%) of the CI children were born in the US.

Table 1.

Means and SDs of Mother Variables at Time 1.

Mother Variables	Chinese Immigrant (n = 60)		European American (n = 71)
Mother Variables	M (SD)	Range	M (SD)	Range
Education	2.65 (0.52)	01–03	2.52(0.63)	01–03
Time (on task in minutes)	7.3 (2.79)	3.34–15.35	7.09(3.25)	2.03–18.52
Mental state language***	10.25 (7.12)	01–32	29.76(18.11)	02–74
Behaviors***	49.23 (20.45)	13–134	36.68(14.60)	10–83

Note. There were significant group differences with regard to references to mental states: t(128) = 7.78, p < 0.001, 95% CI [14.55, 24.46] and behaviors: t(128) = −4.04, p < 0.001, 95% CI [−18.71, −6.42].

CI: Chinese immigrant; M: mean; SD: standard deviation.

***: p < 0.001.

Procedure

Two female researchers visited mothers and children in their homes. Chinese-English bilingual researchers visited the CI families. The children were interviewed in the language with which they were most comfortable. During each visit, before commencing with the interview, the researchers established familiarity and rapport with the children by chatting about non-relevant events. All materials were written in both English and Chinese, and translation and back-translation procedures were carried out to ensure their equivalence in both literal and sense meaning. Mothers were asked to engage the child in a series of free play and semi-structured tasks, followed by a researcher–child session. The entire home visit took approximately 1.5 h and was recorded on videotape.

Measures

Maternal MSL. Mothers’ use of MSL and references to behaviors were assessed in a storytelling task at Time 1. Joint picture-book reading is a common practice among middle-class families in both EA and Chinese cultures (Johnston & Wong, 2002). Mothers received a book entitled “Bear Goes to the Market,” a children’s book with illustrations but no words, 17 pages long. This storybook has been developed for cross-cultural research, and it has been used in previous research with EA and Chinese samples (Han, Leichtman, & Wang, 1998; Wang, Leichtman, & Davies, 2000). The story depicted a mother bear and a baby bear dropping off mail, playing in a playground, and going to the grocery store (where the baby bear briefly loses his mother and gets upset, because he wants a cake and is not allowed to have it). Mothers were asked to tell the story with their children and to make up the text as they went along. They were also told that they could tell the story in any way they would like and could take as long as they wanted to complete the task. For CI mother-child dyads, time spent on the storytelling task ranged from 3.35 to 15.35 min (M = 7.29 min, SD = 2.79). For EA mother-child dyads, it ranged from 2.03 to 18.52 min (M = 7.09 min, SD = 3.25). There was no difference in the mean length of the storytelling between the two cultural groups.

Coding of the story narratives was done in the original language. A native English speaker coded the EA data, and a native Chinese speaker coded the CI data. Repeated training sessions were held to ensure that both coders were applying the same definitions of the coding scheme to the two datasets. A third coder, who was bilingual and fluent in both Chinese and English, independently coded 20% of each dataset for reliability. Any disagreements were resolved through discussion.

The coding scheme was adapted from Ruffman, Slade, & Crowe (2002). Mothers’ references to cognitions (e.g., think, know), desires (e.g., want, like, love), modulations of assertions (e.g., might, maybe), emotional states (e.g., sad, pleased), and other mental states (e.g., remember, decide) were separately tabulated. These references were then combined to form a composite category of the MSL. Coding for the CI data was further informed by research examining MSL in Chinese children (Lu, Su, & Wang, 2008; Tardif & Wellman, 2000). Terms were counted as referring to mental states when the usage was judged, in context, to be indicative of desires (e.g., 要yao4), emotions (e.g., 生气sheng1qi4), cognitions (e.g., 想xiang3, 知道zhi1dao4), other mental states (e.g., 记得ji4de), and modulations of assertions (e.g., 可能ke2neng2). Terms that could be interpreted as behavior or mental states (e.g., 看kan4, equivalent to “see” in English, but may be understood as “think” in Chinese) were coded based on the context. Inter-coder reliability, as indexed by Cohen’s kappas, ranged from 0.96 to 1 for the CI data (M = 0.98) and from 0.87 to 0.96 for the EA data (M = 0.93). A total MSL score was computed for each participant. We used an overall score to be consistent with past research (Doan & Wang, 2010; Ruffman, Slade & Crowe, 2002; Tarullo, Youssef, Frenn, et al., 2016) and because understanding situations that elicit emotions involve understanding of not simply emotions, but desires, and thoughts as well. Moreover, a total score better captures the theoretical notion that individual differences in the frequency of which an individual refers to mental states capture the proclivity of the individual to highlight and emphasize internal or external predictors of behaviors than scores focusing on sub-categories.

Mothers’ references to overt behaviors, including physical actions depicted in the book (e.g., running, playing) and potential actions that were not necessarily seen in the book (e.g., he’s going to go home now) were coded as behavioral descriptions. Cohen’s kappa was 0.83 for the CI data and 0.95 for the EA data. This variable further included a subcategory of mothers’ references to the behavioral/descriptive manifestations of emotions (e.g., “The bear is crying” and “aw, the bear has tears on his face”). Compared to references to internal emotional states (e.g., “The bear is sad”), behavioral descriptions of emotions focused on external or physical aspects of emotions (Ruffman, Slade, & Crowe, 2002). Cohen’s kappa for this sub-category was 0.98 for the CI data and 0.87 for the EA data.

Language ability. Mothers filled out a shortened version of the Child Development Inventory (Ireton, 1992) that assessed children’s language skills at each age. Both expressive language and language comprehension were assessed (possible score range 0–100; Cronbach’s α = 0.93). The questionnaire was translated and back-translated by two English–Chinese bilingual research assistants and checked by a native Chinese speaker to ensure a balance of literal and sense meaning, as well as natural-sounding expression. English words and grammatical rules that were not used in Chinese were exchanged for the Chinese equivalents. The inventory was then pilot-tested with seven EA families and five CI families to ensure equivalence. Because children’s expressive language score and language comprehension scores were highly correlated r = 0.80, p < 0.0001, they were combined for later analysis. The language measure was included because language ability has been shown to be related to the understanding of emotions and mental states (Happé, 1994).

Children’s ESK. Children’s ESK was assessed at three timepoints (see Figure 1), using an emotion production task adapted from previous cross-cultural research examining children’s ability to understand situations that provoke various emotions (Harris, Olthof, Terwogt, et al., 1987). This task has been used with EA and Chinese preschoolers (Wang, Hutt, Kulkofsky, et al., 2006; Wang, 2008) and has been shown to be effective in eliciting children’s responses regarding the situational antecedents of emotions. Children were asked to describe situations likely to provoke fearful, sad, angry, and happy emotions. The researcher presented each term to the child and asked the child to describe situations that would elicit such an emotion in people (e.g., “What makes people feel sad?”) and in the child (e.g., “What makes you feel sad?”). For each question, the researcher prompted the child to provide as many situations as possible (e.g., “What else makes people feel sad?”) until the child indicated that he or she was finished. The researcher rewarded children with stickers to motivate them to come up with as many situations as possible. This task took approximately 10 min.

Figure 1.

Changes in Children’s ESK Over Time.

The number of responses for self and other that children provided were tallied. Although repetition within one category was not counted, if the same response was correct for both the self and for others, it was tabulated. Responses were considered correct if the situation described by the child was judged to be able to elicit the presented emotion (e.g., “What makes you feel happy?” “My teddy bear”; “getting a toy,” “having ice cream”). This included situations that would elicit the emotion in a child, but not necessarily in an adult. Children’s correct responses for each of the four emotions were tallied and summed to form a composite score of ESK. A second research assistant coded 20% of the data for reliability. The average kappa reliabilities were 0.87 (0.71–1.00) for CI children and 0.87 (0.66–1.00) for EA children.

Analytic Strategy

First preliminary analyses were conducted to examine relations and group differences among language, gender, maternal education, and time spent on the book reading task. To examine the relations between mothers’ MSL and children’s trajectories of ESK, latent growth curve (LGC) models (Duncan, Duncan, & Strycker, 2013) were estimated in Mplus 7.5 (Muthén & Muthén, 1998–2007), using the maximum likelihood estimation with robust SE, to accommodate for non-normality and missing data. The FIML estimation procedure includes data from individuals missing on dependent, but not predictor variables (Muthén & Shedden, 1999). For our growth models, the final sample size using FIML was 104. Multi-group latent growth analyses were used to examine group differences.

Results

Preliminary Analyses

Language scores (Ms = 77.38, 86.43, 94.61, SDs = 15.11, 11.87, 6.73) were significantly correlated with ESK scores at Time 1 and Time 2 (age 3: r = 0.27, p = 0.002, 95% CI [0.10, 0.42]; age 3.5: r = 0.27, p = 0.002, 95% CI [0.10, 0.43]). Age and gender were not associated with children’s ESK; however, females had higher levels of language ability at all three timepoints (age 3: t(128) = 2.61, p = 0.01, 95% CI [1.44, 11.67]; age 3.5: t(121) = 2.99, p = 0.003, 95% CI [2.10, 10.32]; age 4.5: t(104) = 2.05, p = 0.042, 95% CI [0.09, 5.13]). Language and gender thus served as covariates for all analyses. Mothers’ education and time spent on task showed no significant effects in preliminary analyses; thus, these variables were no longer considered in further analyses. At T1 (3 years of age), there were 60 CI children and 71 EA children. At T2 (3.5 years of age), there were 48 CI children and 65 EA, and at T3 (4.5 years of age), the sample consisted of 43 CI children and 63 EA children. The majority of children who were not available for later assessments was due to family moves. There were no significant differences in MSL, education, or language abilities between children who remained in the study and those who did not.

We were interested in the sheer amount of exposure to mental state terms, and previous research on mother–child interaction has shown that compared with proportion, frequencies are more predictive of child outcomes (e.g., Fivush, 1998; Han, Leichtman, & Wang, 1998; Reese, Haden, & Fivush, 1993). Additionally, because each utterance potentially has a direct bearing on children’s understanding of mental states because it provides additional input (Ruffman, Slade, & Crowe, 2002), to be consistent with previous research (Doan & Wang, 2010), we used frequencies in our analyses.

To ensure that the number of maternal references to mental states and behaviors did not simply reflect overall maternal involvement or verbosity and, hence, inadvertently contribute to children’s ESK, we examined the sum of mental states and behavioral references (M = 63.28, SD = 27.08) as well as the total amount of time spent on the storytelling task (M = 7.18, SD = 3.04), in relation to children’s ESK. Results revealed no significant correlation between the sum of references and ESK. Similarly, there was no significant relation between the total time spent on the task and children’s ESK at any of the timepoints, thus ruling out the possibility that the overall linguistic input was influencing children’s ESK. In addition, there was no significant cultural difference in mothers’ total references to mental states and behaviors combined, nor in the amount of time mother–child pairs spent on the storytelling task.

Descriptive and Inter-Correlations Across Cultures

Table 1 illustrates the mean and SDs for mother variables and Table 2 illustrates child variables across all three timepoints. There were significant cultural differences in MSL t(128) = 7.79, p < 0.001, 95% CI [14.55, 24.46] and behaviors t(128) = 4.04, p < 0.001, 95% CI [−18.71, −6.42]. EA mothers used significantly more MSL, but made less references to behaviors as compared to CI mothers. Cultural differences were also found in children’s ESK scores at 3, t(129) = 7.96, p < 0.00, 95% CI [7.59, 12.60], and 3.5, t(120) = 4.15, p < 0.001, 95% CI [4.09, 11.55]. EA children had higher levels of ESK at 3 and 3.5. At 4.5 years of age, the culture difference in children’s ESK was no longer significant t(103) = −1.80, p = 0.075. Next, zero-order correlations were calculated between MSL, references to behaviors and children’s ESK at each of the timepoints (Table 3). Across the whole sample, MSL was positively associated with children’s ESK at age 3, r(130) = 0.43, p < 0.001, 95% CI [0.27, 0.57] and 3.5. r(121) = 0.30, p = 0.001, CI [0.12, 0.45], but not at 4.5 r(104) = 0.12, p = 0.24, 95% CI [−0.08, 0.30], whereas references to behavior were negatively associated with ESK at age 3, r(130) = −0.25, p = 0.004, 95% CI [−0.41, −0.08], and age 3.5, r(104) = 0.12, p = 0.24, 95% CI [−0.08, 0.30], but not at 4, r(104) = −0.16, p = 0.11. When correlations were examined across groups (Table 4), for EA children, MSL was positively correlated with references to behaviors r(71) = 0.54, p < 0.001, 95% CI [0.35, 0.68]. Language at 3, r(65) = 0.29, p = 0.02, 95% CI [0.05, 0.49] and 3.5, r (65) = 0.26, p = 0.04 95% CI [0.02, 0.05] were positively related to ESK at 3.5. Language at 4.5 was also correlated with ESK at 4.5, r(62) = 0.33, p = 0.008, 95% CI [0.09, 0.54]. For CI children, MSL was positively correlated with references to behaviors, r(59) = 0.41, p = < 0.001, 95% CI [0.17, 0.60] and ESK at 4.5, r(42) = 0.34, p = 0.03, 95% CI [0.04, 0.58].

Table 2.

Descriptives for Child Variables at all Three Timepoints.

Child Variables	Chinese Immigrants			European Americans
Child Variables	Time 1 (n = 60) M (SD)	Time 2 (n = 57) M (SD)	Time 3 (n = 43) M (SD)	Time 1 (n = 71) M (SD)	Time 2 (n = 65) M (SD)	Time 3 (n = 62) M (SD)
Age in Months	35 (3.43) 30–43	41.46 (3.70) 34–50	54.30 (4.14) 44–63	35.49 (3.30) 30–42	41.89 (2.99) 63–48	53.80 (3.37) 48–61
Language	71.15 (16.44)** 26–99	83.13 (13.20)** 52–100	93.80 (7.9) 65–100	82.48 (11.64) 48–99	89.4 (9.7) 50–100	95.2 (5.6) 69–100
ESK	3.77 (5.19)*** 0–19	9.12 (7.25)** 0–35	17.65 (14.1) 0–69	13.86 (8.58) 0–33	16.94 (12.49) 0–60	22.45 (12.96) 2–60

Note. Significant group differences for language at Time 1 t(129) = 4.60, p < 0.001, 95% CI [6.46, 16.20], and Time 2, t(121) = 2.99, p = 0.003 95% CI [2.12, 10.34] were found. Similarly, there were group differences in ESK at Time 1, t(129) = 7.96, p = 0.001, 95% CI [7.58, 12.60] and Time 2, t(120) = 4.15, p = 0.001, 95% CI [4.09, 11.54].

CI: Chinese immigrant; ESK: emotion situation knowledge; M: mean; SD: standard deviation.

The range is reported below each mean. We tested for group differences at each timepoint. Asterisks represent the following significance levels: **p < 0.01, ***p < 0.001.

Table 3.

Zero-Order Correlations among Mother and Child Variables for Whole Sample.

	1	2	3	4	5	6	7	8
1. Mental state language	—	0.14	0.27***	0.21*	0.03	0.43**	0.30**	0.030
2. Behaviors		—	-0.01	−0.07	−0.03	-0.25**	−0.24**	−0.16
3. Language (age 3)			—	0.83***	0.56***	0.27**	0.31**	0.15
4. Language (age 3.5)				—	0.66***	0.22*	0.27**	0.21*
5. Language (age 4.5)					—	0.11	0.18⁺	0.17⁺
6. Emotion situation knowledge (age 3)						—	0.53***	0.32**
7. Emotion situation knowledge (age 3.5)							—	0.34***
8. Emotion situation knowledge (age 4.5)								—

Note. Sample sizes were n = 131, 123, and 105 at times 1, 2, and 3, respectively.

*p < 0.05; **p < 0.01; ***p < 0.001 (all significance tests are two-tailed).

Table 4.

Zero-Order Correlations among Mother and Child Variables by Ethnicity.

	1	2	3	4	5	6	7	8
European Americans (above the diagonal)
1. Mental State Language	−	0.54***	0.07	0.06	0.03	0.20	0.15	−0.05
2. Behaviors	0.41**	−	0.14	0.20	0.07	−0.10	−0.15	−0.21
3. Language (age 3)	0.13	0.13	−	0.81**	0.62***	0.03	0.29*	0.09
4. Language (age 3.5)	0.18	0.17	0.81***	−	0.71***	0.00	0.26*	0.20
5. Language (age 4.5)	−0.16	−0.05	0.52**	0.61***	−	0.02	.20	0.33**
6. Emotion Situation Knowledge (age 3)	−0.03	−0.06	0.16	0.24	0.16	−	0.48***	0
7. Emotion Situation Knowledge (age 3.5)	0.04	−0.12	0.12	0.15	0.11	0.25*	−	0.34**
8. Emotion Situation Knowledge (age 4.5)	0.34*	0.00	0.12	0.16	0.00	0.23	0.23	−
Chinese Immigrants (below the diagonal)

Note. European Americans (n = 71, 65, and 62, at age 3, 3.5, and 4.5, respectively) are above the diagonal and Chinese immigrants (n = 60, 58, 43) are below the diagonal.

*p < 0.05; **p < 0.01; ***p < 0.001 (all significance tests are two-tailed).

Multi-Group Unconditional Models

An LGC analysis was used to investigate the change in children’s ESK and to test whether growth in children’s ESK was related to mothers’ discourse variables, and to see if the effects of the discourse variables varied across cultures. A baseline unconditional LGC model of children’s ESK was fit from children’s scores at each timepoint. The time scores for the slope growth factors were fixed at 0, 0.5, and 1.5 (reflecting time of the initial data collection, then half a year, and 1.5 years later). The multi-group unconditional growth model suggested excellent overall fit χ² = (2, N = 104) = 0.57, p = 0.75, comparative fit index (CFI) = 1.00, root-mean-square error of approximation (RMSEA) = 0.00, 95% CI [0.00, 0.12]. However, a Heywood case was observed in the CI group (a non-significant correlation between the intercept and the slope, which is likely due to the small sample size). To resolve the issue, we constrained the non-significant correlation between the intercept and the slope for both groups to be 0. The model converged well, the overall fit was excellent, χ² = (4, N = 104) = 2.16, p = 0.71, CFI = 1.00, RMSEA = 0.00, 95% CI [0.00, 0.21]. In addition, fit statistics also suggest that the model fit well for both EA χ² = (2, N = 130) = 1.28, p = 0.53, CFI = 1.00, RMSEA = 0.00, 95% CI [0.00, 0.21] and CI participants χ² = (2, N = 130) = 0.86, p = 0.65, CFI = 1.00, RMSEA = 0.00, 95% CI [0.00, 0.20]. Additional models were conducted where we systematically constrained residual variances and factor variances. The difference in the covariance between the intercept and the slope was not tested because the variance of the intercept was not equal across groups. The best fitting multi-group unconditional model allowed the residual variance at time 2, and the variance in the intercept to vary across groups, χ² = (2, N = 130) = 0.86, p = 0.65, CFI = 1.00, RMSEA = 0.00, 95% CI [0.00, 0.20]. This model is the basis for the conditional models. In addition, constraining both the intercept and the slope factor means lead to significantly worse model fit (Δχ² (1) = 8.8, Δp = 0.003; (Δχ² (1) = 96.29, Δp = 0.9.9, respectively), suggesting cultural differences in both the intercept and slope.

Multi-Group Conditional Models

Conditional LGC models were fit with the data to test whether the between-person variation in growth parameters in the unconditional baseline model was related to variation in the predictors. Maternal references to behaviors and mental states were included as predictors of the intercept and the slope; sex served as a time-invariant covariate and language was included as a time-varying covariate. Equal constraints from the unconditional models were carried forward to the conditional models, but otherwise all parameters were allowed to vary across groups. This model fit the data well χ² = (25, N = 104) = 21.50, p = 0.66, CFI = 1.00, TLI = 1.10, RMSEA = 0.00 95% CI [0.00, 0.09].

The chi-squared contribution from each group was 13.18, and 8.32 for EAs and CIs, respectively. For EA children, MSL was positively associated with the intercept, B = 0.18, SE = 0.06, p = 0.006, 95% CI [0.05, 0.30], whereas behavior had a negative association, B = −0.20, SE = 0.07, p = 0.006, 95% CI [−0.34, −0.06]. None of the variables predicted the slope. Language at 3.5 years (B = 0.31, SE = 0.11, p = 0.007, 95% CI [0.09, 0.54]), and 4.5 years (B = 0.86, SE = 0.24, p < 0.001, 95% CI [0.40, 1.34]), were also significantly associated with children’s ESK at 3.5 and 4.5, respectively. For CI children, references to behaviors was negatively associated with the intercept (B = − 0.07, SE = 0.03, p = 0.03, 95% CI [−0.14, −0.01]. MSL was significantly associated with the slope of ESK (B = 0.52, SE = 0.26, p = 0.04, 95% CI [0.02, 1.04]. Next, we test the extent to which the association between the predictors on the slope varied across group. Because intercept variance is different across groups, meaningful comparisons cannot be made on the intercept (Duncan, Duncan, & Strycker, 2013).

We first set the regression parameters to be equal across groups. This resulted in significantly worse model fit, χ² = (28, N = 104) = 33.45, p = 0.22, CFI = 0.91, TLI = 0.86, RMSEA = 0.06 95% CI [0.00, 0.13]. The Δχ² between the constrained and unconstrained model was significant Δχ² = 11.18, Δdf = 3, Δp = 0.01. We then systematically constrained each of the regression coefficients. Constraining the path of gender and references behaviors to the slope did not lead to significant differences in fit. However, constraining the effect of MSL on the slope led to significantly worse fit, ▵χ² = 13.15, Δdf = 1, Δp < 0.001, suggesting that the effect of MSL should be allowed to vary across groups. Fit for the final model was excellent, χ² = (27, N = 104) = 21.55, p = 0.76, CFI = 1.00, TLI = 1.15, RMSEA = 0.00 95% CI [0.00, 0.08]. Figure 2 depicts the final model with estimates for each group.

Figure 2.

Estimates and SEs for the Final Model.

Discussion

Understanding emotions and their situational antecedents are crucial for children to navigate their social worlds and to regulate their own emotions. However, culture impacts the socialization processes that affect children’s emotion understanding. This study examines mothers’ use of MSL and behaviors and children’s ESK in a prospective, cross-cultural context. Findings from the LGC models demonstrated that there are significant cultural differences with regard to both emotional understanding at age 3 and the trajectories of emotional understanding over time. EA children started with higher emotional understanding than their CI counterparts. However, CI children grew faster over time, such that by 4.5 there were no cultural differences in children’s ESK. CI mothers also made more references to behaviors and fewer references to mental states as compared to EA mothers. Mothers’ references to MSL were associated with children’s ESK at age three for EA children. However, when examining the growth of ESK, mothers’ use of MSL was positively associated with the slope of CI children’s ESK. Mothers’ references to behaviors were negatively associated with children’s ESK at three years of age for both groups.

Consistent with the theoretical notion that individual realities, emotions, and thoughts may not be as salient in interdependent cultures as in independent cultures (Markus & Kitayama, 1991; Potter, 1988), we found that Chinese mothers made fewer references to mental states during book reading than EA mothers made. On the other hand, given that individuals in interdependent cultures are often characterized by their external behaviors rather than by inner thoughts and feelings, CI mothers made more references to behaviors. These differential maternal narrative inputs, in turn, are associated with children’s ESK. In particular, MSL may draw children’s attention to the psychology of individuals and give children a vocabulary to think and discuss these concepts, which may facilitate the development of emotion understanding. On the other hand, references to behavior may focus the child instead onto external facets, encouraging the child to focus on the action and perhaps its consequences. At the same time, they may also highlight the importance of one’s behaviors on the outside world. This sensitivity to context and effects on others are consistent with an interdependent worldview.

References to mental states may be particularly facilitative for the development of emotion knowledge during the period when children’s expression and understanding of mental states are beginning to emerge (Jenkins, Turrell, Kogushi, et al., 2003). In our data, mothers’ references to MSL were associated with concurrent ESK for EA children, but not the slope growth factor. Note, however, that we did not capture how mothers’ were using MSL, nor did we assess maternal references to MSL at later timepoints. It is possible that if we looked in detail at mental state references usage, for example, in the context of explanations or elaborations, the results might be different. Indeed, one possibility for lack of consistency regarding the longitudinal relations between maternal MSL and children’s social cognition may be that as children advance in their social cognitive abilities, simple references to mental states may become less beneficial whereas more complex linguistic inputs such as elaborations and explanations may become more vital (e.g., Ontai & Thompson, 2008). Alternatively, it is also possible that as children age, family factors may become less important, and children’s emotion knowledge may be more influenced by peers, or intrinsic child factors such as language competencies, or previously established trajectories (Jenkins, Turrell, Kogushi, et al., 2003). References to mental states were associated with changes in ESK for CI children, suggesting that timing of the importance of MSL varies across cultures.

Our data suggest that future research should consider the complexity of mother–child discourse, specifically discourse that moves beyond mere references to internal states to elaboration and explicit explanations of causal relationships between psychological states and behaviors. As a case in point, Ontai & Thompson (2008) found that in the context of mother–child memory sharing, maternal elaborations, or statements that move the conversation to a new aspect or adds more details, predicted theory of mind, whereas maternal references to mental states showed no effect. Similarly, in examining both theory of mind and emotion knowledge, La Bounty, Wellman, Olson, et al. (2008) found a similar pattern of results for fathers’ causal talk, although it did not reach significance when examined longitudinally. Future research examining children’s emerging social cognitive abilities would benefit greatly by looking specifically at both the content and the style of mother–child discourse in relation to the child’s developmental age. By examining interactions between types of maternal discourse style and children’s development, we would have a better understanding of the type of discourse that matters, and when they matter in children’s development.

Important strengths of this study are the cross-cultural sample, and the measurement of ESK at multiple timepoints, allowing us to investigate prospective relations between MSL and changes in ESK over time. However, in this study, we only measured references to MSL at one timepoint and were thus unable to assess whether mothers’ use of MSL increased over time or how this development would parallel the development of children’s ESK. Although maternal references to MSL were not associated with trajectories of EA children’s ESK, concurrent or later references may continue to be of importance. Nevertheless, research has shown that mothers tend to be consistent in their MSL over time (Fivush & Baker-Ward, 2005). Also, this ESK measure relies heavily on children’s ability to produce language, rather than nonverbal indicators such as pointing, and thus may have resulted in a more conservative assessment of their cognitive abilities.

Finally, in future studies, it would be informative to see whether the effects of acculturation on CI mothers would influence their use of mental state terms. A study of Pacific Island families living in New Zealand by Taumoepeau (2015) showed that parents’ ethnic identity predicted differential growth trajectory of children’s performances in emotion situation and knowledge tasks. Theoretically, as immigrants become more acculturated with a host culture and began to adopt the cultural norms and values, they would be more likely to adopt the cultural behaviors as well. Additionally, research examining cross-cultural differences in emphasis on mental states versus behaviors would benefit greatly by measuring the variable of interest for which culture is acting as a proxy. For example, within a culture, are those who score high on interdependence measures less likely to use MSL? Would this depend on the discourse context (e.g., talking with a person of the same or different ethnic group)?

Furthermore, despite including longitudinal analyses, our data are still observational. Thus causal direction must be interpreted with caution. It is possible that children with higher ESK are more likely to elicit more references to mental states from their mothers. However, this is unlikely, given that research has shown that mothers’ use of MSL is not influenced by child characteristics (Garrett-Peters, Mills-Koonce, Adkins, et al., 2008). Additionally, our analyses also did not include children’s MSL as a predictor of ESK due to its very low occurrence in our sample. We did control for children’s language capacity when looking at trajectories of change. Finally, experimental studies that manipulated maternal mental state input found that references to MSL improved children’s theory of mind understanding (Guajardo & Watson, 2002). These studies suggest that the result of children’s understanding of mental states is not solely child-driven.

In sum, findings from this study suggest that it is important to consider the developmental level of the child, in specific cultural contexts, to understand the effects of MSL on children’s developing emotion understanding. Although we do not presume a universal perspective on development, we focused on MSL research, which has its origins in Western countries. Other individual or socialization factors not captured in this study may be more important for understanding children’s social–cognitive development across cultures. For example, individual characteristics such as executive function, or other ways of discussing emotions (highlighting behaviors and consequences rather than internal states), may offer qualitatively different developmental pathways to the same end point. A cross-cultural approach is critical to enriching our theoretical understanding of social–cognitive mechanisms underlying child development.

Footnotes

Acknowledgment

We thank members of the Social Cognition Development Lab at Cornell University for their contributions to the project.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by NIMH Grant R01-MH64661 to Qi Wang.

ORCID iD

Stacey N. Doan

References

Bartsch

Wellman

H.M.

(1995). Children Talk about the Mind. Oxford, UK: Oxford University Press.

Bloom

(2000). How Children Learn the Meanings of Words. Cambridge, MA: MIT Press.

Carpendale

Lewis

(2004). Constructing an understanding of mind: The development of children’s social understanding within social interaction. Behavioral and Brain Sciences, 27, 79–151. doi:10.1017/S0140525X04000032.

Clancy

P. M.

(1999). The socialization of affect in Japanese mother-child conversation. Journal of Pragmatics, 31, 1397–1421.

Denham

S. A.

(1993). Maternal emotional responsiveness and toddlers’ social-emotional competence. Journal of Child Psychology and Psychiatry, 34, 715–728. doi:10.1111/j.1469-7610.1993.tb01066.x.

Denham

Zoller

Couchod

(1994). Socialization of preschoolers’ emotion understanding. Developmental Psychology, 30, 928–936. doi:10.1037/0012-1649.30.6.928.

Doan

S. N.

Wang

(2010). Maternal discussions of mental states and behaviors: Relations to emotion situation knowledge in European American and immigrant Chinese children. Child Development, 81, 1490–1503.

Duncan

T. E.

Duncan

S. C.

Strycker

L. A.

(2013). An Introduction to Latent Variable Growth Curve Modeling: Concepts, Issues, and Application. Delhi, India: Routledge Academic.

Dunn

Bretherton

Munn

(1987). Conversations about feeling states between mothers and their young children. Developmental Psychology, 23, 132–139. doi:10.1037/0012-1649.23.1.132.

10.

Dunn

Brown

Beardsall

(1991). Family talk about feeling states and children’s later understanding of others’ emotions. Developmental Psychology, 27, 448–455. doi: 10.1037/0012-1649.27.3.448.

11.

Dunn

Brown

Slomkowski

Tesla

Youngblade

(1991). Young children’s understanding of other people’s feelings and beliefs: Individual differences and their antecedents. Child Development, 62, 1352–1366. doi:10.1111/j.1467-8624.1991.tb01610.x.

12.

Eisenberg

Cumberland

Spinrad

(1998). Parental socialization of emotion. Psychological Inquiry, 9, 241–273. doi:10.1207/s15327965pli0904_1.

13.

Fivush

(1998). Interest, gender and personal narrative: How children construct self-understanding. In Karp

Renninger

Baumeister

Hoffman

(Eds.), Interest and Gender in Education. (pp. 58–73). Kiel, Germany: Institute for Science Education.

14.

Fivush

Baker-Ward

(2005). The search for meaning: Developmental perspectives on internal state language in autobiographical memory. Journal of Cognition and Development, 6, 455–462. doi:10.1207/s15327647jcd0604_1.

15.

Fivush

Wang

(2005). Emotion talk in mother-child conversations of the shared past: The effects of culture, gender, and event valence. Journal of Cognition and Development, 6, 489–506. doi:10.1207/s15327647jcd0604_3.

16.

Garner

P. W.

Jones

D. C.

Gaddy

Rennie

K. M.

(1997). Low-income mothers’ conversations about emotions and their children’s emotional competence. Social Development, 6, 37–52. doi:10.1111/j.1467-9507.1997.tb00093.x.

17.

Garrett-Peters

Mills-Koonce

Adkins

Vernon-Feagans

Cox

(2008). Early environmental correlates of maternal emotion talk. Parenting, 8, 117–152. doi:10.1080/15295190802058900.

18.

Guajardo

N. R.

Watson

A. C.

(2002). Narrative discourse and theory of mind development. The Journal of Genetic Psychology, 163, 305–325. doi:10.1080/00221320209598686.

19.

Han

J. J.

Leichtman

M. D.

Wang

(1998). Autobiographical memory in Korean, Chinese, and American children. Developmental Psychology, 34, 701–713. doi:10.1037//0012-1649.34.4.701.

20.

Happé

F. G.

(1994). An advanced test of theory of mind: Understanding of story characters’ thoughts and feelings by able autistic, mentally handicapped and normal children and adults. Journal of Autism and Developmental Disorders, 24, 129–154. doi:10.1007/BF02172093.

21.

Harris

P. L.

(1989). Children and Emotion. New York, NY: Cambridge University Press.

22.

Harris

P. L.

Olthof

Terwogt

Hardman

(1987). Children’s knowledge of the situations that provoke emotion. International Journal of Behavioral Development, 10, 319–343. doi:10.1177/016502548701000304.

23.

Hofmann

S. G.

Doan

S. N.

(2018). The Social Foundations of Emotions. Washington, DC: American Psychological Association.

24.

Ireton

(1992). Child Development Inventory. Minneapolis, MN: Behavioral Science Systems.

25.

Jenkins

Turrell

Kogushi

Lollis

Ross

(2003). A longitudinal investigation of the dynamics of mental state talk in families. Child Development, 74, 905–920. doi:10.1111/1467-8624.00575.

26.

Johnston

Wong

(2002). Cultural differences in beliefs and practices concerning talk to children. Journal of Speech, Language, and Hearing Research, 45, 916–926. doi:10.1044/1092-4388(2002/074).

27.

La Bounty

Wellman

H. M.

Olson

Lagattuta

K. H.

Liu

(2008). Mothers’ and fathers’ use of internal state language. Social Development, 17, 757–775. doi:10.1111/j.1467-9507.2007.00450.x.

28.

Wang

(2008). Talking about others facilitates theory of mind in Chinese preschoolers. Developmental Psychology, 44, 1726–1736. doi:10.1037/a0013074.

29.

Markus

Kitayama

(1991). Culture and the self: Implications for cognition, emotion, and motivation. Psychological Review, 98, 224–253. doi:10.1037/0033-295X.98.2.224.

30.

Mascolo

M. F.

(2004). The coactive construction of selves in cultures. In Mascolo

M. F.

(Eds.), Culture and developing selves: Beyond dichotomization. New Directions in Child and Adolescent Development Series. Damon

(Series Editor). San Francisco, CA: Jossey-Bass.

31.

Mcquaid

Bigelow

A. E.

McLaughlin

MacLean

(2008). Maternal mental state language and preschool children’s attachment security: Relation to children’s mental state language and expressions of emotional understanding. Social Development, 17, 61–83.

32.

Muthén

(1998–2007). Mplus Software. The Comprehensive Modeling Program for Applied Researchers User’s Guide. Los Angeles, CA: Muthén & Muthén.

33.

Muthén

Shedden

(1999). Finite mixture modeling with mixture outcomes using the EM algorithm. Biometrics, 55, 463–469.

34.

Ochs

Schieffelin

(1989). Language has a heart. Text-Interdisciplinary Journal for the Study of Discourse, 9, 7–26.

35.

Ontai

L.L.

Thompson

R.A.

(2002). Patterns of attachment and maternal discourse effects on children’s emotion understanding from 3 to 5 years of age. Social Development, 11, 433–450. doi:10.1111/1467-9507.00209.

36.

Ornstein

P. A.

Haden

C. A.

Hedrick

A. M.

(2004). Learning to remember: Social-communicative exchanges and the development of children’s memory skills. Developmental Review, 24, 374–395. doi:10.1016/j.dr.2004.08.004.

37.

Potter

S. H.

(1988). The cultural construction of emotion in rural Chinese social life. Ethos, 16, 181–208. doi:10.1525/eth.1988.16.2.02a00050.

38.

Preacher

K. J.

Wichman

A.L.

MacCallum

Briggs

N.E.

(2008). Latent Growth Curve Modeling. Thousand Oaks, CA: Sage Publications.

39.

Reese

Haden

Fivush

(1993). Mother–child conversations about the past: Relationships of style and memory over time. Cognitive Development, 8, 403–430. doi:10.1016/S0885-2014(05)80002-4.

40.

Rudek

D. J.

Haden

C. A.

(2005). Mother’s and preschooler’s mental state language during reminiscing over time. Merrill-Palmer Quarterly, 51, 523–549. doi:10.1353/mpq.2005.002.

41.

Ruffman

Slade

Crowe

(2002). The relation between children’s and mother’s mental state language and theory-of-mind understanding. Child Development, 73, 734–751. doi:10.1111/1467-8624.00435.

42.

Symons

(2004). Mental state discourse, theory of mind, and the internalization of self–other understanding. Developmental Review, 24, 159–188. doi:10.1016/j.dr.2004.03.001.

43.

Tardif

Wellman

(2000). Acquisition of mental state language in Mandarin and Cantonese speaking children. Developmental Psychology, 36, 25–43. doi:10.1037//0012-1649.36.1.25.

44.

Tarullo

A. R.

Youssef

Frenn

K. A.

Wiik

Garvin

M. C.

Gunnar

M. R.

(2016). Emotion understanding, parent mental state language, and behavior problems in internationally adopted children. Development and Psychopathology, 28, 371–383.

45.

Taumoepeau

(2015). From talk to thought strength of ethnic identity and caregiver mental state talk predict social understanding in preschoolers. Journal of Cross-Cultural Psychology, 46, 1169–1190. doi:10.1177/0022022115604393.

46.

Taumoepeau

Ruffman

(2006). Mother and infant talk about mental states relates to desire language and emotion understanding. Child Development, 77, 465–481. doi:10.1111/j.1467-8624.2006.00882.x.

47.

Taumoepeau

Ruffman

(2008). Stepping stones to other’s minds: Maternal talk relates to child mental state language and emotion understanding at 15, 24, and 33 months. Child Development, 79, 284–302. doi:10.1111/j.1467-8624.2007.01126.x.

48.

Wang

(2001). “Did you have fun?”: American and Chinese mother–child conversations about shared emotional experiences. Cognitive Development, 16, 693–715. doi:10.1016/S0885-2014(01)00055-7.

49.

Wang

(2003). Emotion situation knowledge in American and Chinese preschool children and adults. Cognition & Emotion, 17, 725–746. doi:10.1037/0012-1649.40.1.3.3.

50.

Wang

(2008). Emotion knowledge and autobiographical memory across the preschool years: A cross-cultural longitudinal investigation. Cognition, 108, 117–135. doi:10.1016/j.cognition.2008.02.002.

51.

Wang

Fivush

(2005). Mother-child conversations of emotionally salient events: Exploring the functions of emotional reminiscing in European American and Chinese families. Social Development, 14, 473–495. doi:10.1111/j.1467-9507.2005.00312.x.

52.

Wang

Hutt

Kulkofsky

McDermott

Wei

(2006). Emotion situation knowledge and autobiographical memory in Chinese, immigrant Chinese, and European American 3-year-olds. Journal of Cognition and Development, 7, 95–118. doi:10.1207/s15327647jcd0701_5.

53.

Wang

Leichtman

M. D.

Davies

K. I.

(2000). Sharing memories and telling stories: American and Chinese mothers and their 3-year-olds. Memory, 8, 159–177. doi:10.1080/096582100387588.

54.

Welch-Ross

M. K.

Fasig

L. G.

Farrar

M. J.

(1999). Predictors of preschoolers’ self-knowledge: Reference to emotion and mental states in mother-child conversation about past events. Cognitive Development, 14, 401–422.

55.

Wellman

H. M.

Harris

P. L.

Banerjee

Sinclair

(1995). Early understanding of emotion: Evidence from natural language. Cognition and Emotion, 9, 117–149. doi:10.1080/02699939508409005.

56.

Wood

Conway

(2006). Subjective impact, meaning making, and current and recalled emotions for self-defining memories. Journal of Personality, 74, 811–846. doi:10.1111/j.1467-6494.2006.00393.x.