Parental Attitudes and Children’s Language Development: A Cross-Cultural Evaluation of the Early Parenting Attitudes Questionnaire Open Access

Culture, as well as ethnicity and socioeconomic status (SES), are linked to parenting practices, which can influence children’s mental health through, for example, parents’ expectations, the behaviors they value, and the type of care they provide (Bornstein, 2013). Parenting combines intuition and knowledge, which are often acquired by living in a culture and might differ across countries. There are differences in cultural and societal norms shaping children’s upbringing, as well as in belief systems regarding, for example, socially acceptable and unacceptable behaviors or the type of support needed for children’s development. For instance, in many countries, it is normative for parents to talk to infants and believe they understand speech long before they start producing speech themselves, but in certain societies, such as the Tsimane of Bolivia, parents do not engage in infant-directed speech very often (Cristia et al., 2019). Parental attitudes are also known to affect the emotional development of children (Eisenberg et al., 2005). As Morris et al. (2007) suggest, children learn about emotion regulation by observation. Moreover, they argue that emotion regulation is affected by parenting practices and by the emotional climate of the family. For instance, parental harsh responses to their children’s negative emotions are linked to lower levels of emotional competence (Jones et al., 2002) and inappropriate emotion regulation strategies (Eisenberg & Fabes, 1994).

Given the cultural variation in parenting attitudes and beliefs, evaluating the cross-cultural validity of the instruments that assess them is crucial. However, disentangling the effect of culture and language is challenging, considering that the three countries of our sample differ not only in terms of culture but also in their respective languages (i.e., Russian, Norwegian, and English). It is important to note here that language is an integral part of culture and in most cases not distinguishable from it. The first two aims of the present research study are to assess the psychometric properties of the EPAQ (Hembacher & Frank, 2020) across three different languages and cultures and to investigate whether the underlying structure aligns with that observed in the U.S. The EPAQ was chosen among other instruments for several reasons, in particular, its recency and the diversity of domains covered by its items. For instance, compared to the Parent/Caregiver Involvement Scale – Short Form (P/CIS-SF; Taylor & Bergin, 2019) and the Survey of Parent/Provider Expectations and Knowledge (SPEAK; Suskind et al., 2018), which are supposed to measure intuitive theories of parenting related to specific domains (i.e., in high-risk environments, and related to cognitive and language development), the EPAQ is broader and assesses more dimensions of parenting. At the same time, compared to the Knowledge of Infant Development Inventory (KIDI; MacPhee, 1981), the EPAQ is less broad and more focused on psychological aspects of parenting.

Several factors contributed to the choice of the three countries of our samples, including practical reasons such as ongoing collaborative work with international teams based in Russia, Norway, and the U.K. However, these countries were primarily chosen because of the differences in their cultural backgrounds. Central and Eastern European societies are more collectivist than Western European and American ones (e.g., Hofstede, 1980; Kolman et al., 2003), potentially due to the different political situations present in these countries. Central and East Europeans experienced a Communist rule relatively recently, which “encouraged collectivist thinking and behavior and structured life in such a way that reliance on others was necessary for survival, and personal connections with others were necessary for success” (Varnum et al., 2008, p. 324). In fact, prior studies found, for instance, that Russians are more collectivist than the British (Tower et al., 1997, as cited in Varnum et al., 2008) and that, compared to West Europeans, Central and East Europeans place more importance on hierarchy and less on autonomy (Schwartz & Bardi, 1997). Despite the social and economic changes that Russia went through, a recent study by Borshchevskiy (2022) found that collectivist values are still dominant in Russian society, with some differences in the high-power distance.

Individualist and collectivist cultures also differ in some aspects of emotions. According to Tsai et al. (2007), people in Western or individualist cultures are encouraged to communicate their innermost thoughts and feelings and to exert influence over others via high-arousal emotions. On the contrary, low-arousal emotions are valued more in Eastern or collectivist cultures, as conforming to others is a desirable behavior (Tsai et al., 2007). Although individualism and collectivism are two of the most studied constructs used to describe Western and Eastern cultures, recent studies suggest that they might be too broad and end up masking some aspects of the societies they describe (Lomas et al., 2023). Moreover, individualism and collectivism might not be the only factors to take into account when considering cross-cultural differences. Indeed, a study by Lin et al. (2017) found that nations that are conventionally regarded as individualistic or collectivistic do not differ much in terms of attachment orientations and psychological outcomes, indicating that there are also other factors involved in the relationship between attachment and psychological outcomes. One of these factors could be, for instance, power distance. The latter is a cultural dimension identified by Hofstede (2011), which is defined as the degree to which the weaker members (e.g., children) of a group (e.g., a family) accept an unequal allocation of power within the group. In low power distance societies (e.g., Norway, the U.K., and the US), parents tend to treat children as equals. In high power distance societies (e.g., Russia), obedience is highly important, and elderly people are respected and feared. The different factor structures that we report later on in the present study further confirm that there are indeed substantial cross-cultural differences in Russia, Norway, and the U.K. when it comes to parenting.

The third aim of the study is to explore the potential relationship between parental attitudes and beliefs and children’s language development. For the latter, we used the MacArthur-Bates Communicative Development Inventories (CDI; Fenson et al., 2007)—parent-reported measures of early language comprehension and production administered in the form of vocabulary checklists.

To achieve these goals, we recruited parents of young children in Russia, Norway, and the U.K. and assessed different aspects of intuitive parenting theories, asking parents to fill in the EPAQ (Hembacher & Frank, 2020) as well as the CDIs (Fenson et al., 2007) in their respective languages. All of the translations and adaptations were completed as part of previous studies. Additional information about the translation procedure can be found in the Methods section.

To address the first two aims, we estimated the measurement invariance of the scale (i.e., whether the same factor structure could be observed between countries – a pre-requisite for cross-cultural comparisons) and compared it to that revealed in the American sample. To address the third aim, we evaluated whether parental attitudes correlated with children’s vocabulary size, as indexed by the CDIs (Fenson et al., 2007). Moreover, given that previous studies found that children from lower SES environments usually have smaller vocabularies compared to those from high SES environments (e.g., Pace et al., 2017; Rowe, 2018), we collected data on maternal education that we used as a proxy for SES in all the analyses to control for its possible impact on vocabulary size.

Our reasoning was the following: If the factor structure and the psychometric properties of the EPAQ observed in the original American sample can be replicated in the U.K., Norway, and Russia, then we would conclude that the underlying phenomenon can be generalized to these countries. Conversely, if the underlying phenomenon is not generalizable, the factor structure would fail to replicate. In this case, we would conclude that implicit parenting theories differ substantially between the countries, indicating that the EPAQ, in its current form, is not suitable for use outside the North American context.

After conducting an initial factor analysis, we pre-registered a set of hypotheses (see https://osf.io/puqj9/?view_only=ffe1c74400f44a6788f51e588df838a7). Specifically, we expected that the first factor identified in Russia, named Adult Authority (i.e., based on respect for the adult, learning, and emotional control), would correlate positively with vocabulary, given that, although these parents set high expectations for their children, they still value independence, which has been shown to build confidence and, thus, get the most out of learning opportunities (Cerino, 2023). By contrast, we assumed that the second factor in Russia, named Communicative and Emotional Detachment (i.e., based on the idea that too much attention can spoil a child), and the third factor, named Confrontation Between Child and Adult (i.e., based on a strict hierarchy), would correlate with lower vocabulary scores. For the Norwegian and the U.K. data, we expected that the factor Affection and Attachment (i.e., based on emotional closeness) would correlate positively with vocabulary, given that the majority of the items that load on this factor concerned the child’s safety and emotional wellbeing. In fact, a child living in a safe environment and developing a secure attachment should generally feel more comfortable speaking and interacting with others (Bowlby, 1969/1982), potentially translating into better vocabulary scores. We also expected that the factor referred to as School Preparation (i.e., based on parents’ involvement in education) would associate positively with vocabulary, as parent involvement in school is associated with positive educational outcomes (e.g., Barnard, 2004; Zellman & Waterman, 1998). However, we expected that the factor Rules and Respect (i.e., based on behavioral control) would correlate with lower vocabulary, given that children living in this context may feel afraid to ask questions and engage in new activities, as they might not be sure whether their parents will approve or not.

Our initial sample consisted of 3333 parents of children from Norway (n = 1060), the U.K. (n = 656), and Russia (n = 1617). The mean age of the children was 730.35 days (range 0 to 156 months¹), with a standard deviation of 351.71. The gender of the children was relatively well-balanced in each country. Table 1 provides an overview of the demographic variables.

Most of the data were collected through online questionnaires. The data collection started in August 2019 in Norway, in March 2020 in the U.K., and in April 2020 in Russia. To reach out to as many people as possible, various means were used to recruit participants (e.g., contacting individuals registered in lab databases, promotion through social media, etc.). Participants were also recruited through email invitations. The data collection took place on Nettskjema (https://nettskjema.no/) in Norway, on Google Forms in the U.K., and on Testograf in Russia (https://www.testograf.ru/). The participant compensation varied across the countries, ranging from illustrated books in Russia to lotteries in the U.K. and Norway. An informed written consent was signed by the parents before participation. Each lab obtained ethical approval from their respective institutions. In Russia, the study and consent procedures were approved by the Ethics Committee of Faculty of Psychology at Lomonosov Moscow State University (approval No. 2020/61) . In Norway, the study was approved by the ethics committee of the Department of Psychology of the University of Oslo. In the U.K., the study was approved by the Oxford Brookes University.

In the first part of the survey, parents were asked to fill in their child’s personal information, including the date of birth, gender, native language, and to provide their family information, such as number of siblings, and the highest level of education achieved by parents. Maternal education was used as a proxy for SES in all the analyses to account for its possible correlation with language development. Given that in the U.K. and Russia, fathers have no or very limited parental leaves, mothers usually take care of their children, and paternal education was not used as a proxy. Afterward, parents were asked to fill in the Early Parenting Attitudes Questionnaire in their respective languages. Finally, parents were provided with the language-specific CDI vocabulary checklists (described below), and they were asked to check the words that their child understands and produces (for infants between 8 and 18 months old) or produces only (for toddlers between 18 and 36 months old).

Early Parenting Attitudes Questionnaire. The Early Parenting Attitudes Questionnaire (EPAQ) by Hembacher and Frank (2020) was used to assess aspects of intuitive parenting theories. The questionnaire includes three scales, with eight items each, that were conceptualized to cover different dimensions of parenting attitudes. It asks parents to rate their agreement or disagreement with each statement on a 7-point Likert scale from 0 (completely disagree), to 6 (completely agree). The first subscale, Affection and Attachment, includes statements about emotions and relationships (e.g., “Children should be comforted when they are scared or unhappy,” or, “It is important for parents to help children deal with their emotions”). The second subscale, Early Learning, contains statements about the potential educational value attributed to activities that the child carries out on their own or with their parents (e.g., “It is good to let children explore and experiment,” or, “Parents can help babies learn language by talking to them”). The third subscale, Rules and Respect, assesses the parents’ attitudes when it comes to controlling their child’s behavior (e.g., “It is very important that children learn to respect adults, such as parents and teachers,” or, “It is okay if children boss around their caregivers”).

The questionnaire was translated from English into Russian by a professional translator who lives in Russia and is familiar with the Russian culture, in accordance with the ITC Guidelines for Translating and Adapting Tests (Muñiz et al., 2013). When it comes to cross-cultural adaptation, there is no clear-cut recommendation about which method works best (Epstein et al., 2015). Moreover, the adaptation process of an instrument may be demanding and challenging, especially when it takes into account constructs that cannot be measured directly, like attitudes (Gjersing et al., 2010). The translation process took into account the differences between the languages while preserving all the original characteristics of the test. To ensure the consistency of the two versions of the test, the first Russian translation was finalized in collaboration with bilingual experts in the field of child development, namely Natalia Kartushina and Margarita Gavrilova. Following this, an expert discussion was held to determine whether further adjustments were necessary. As the written languages of Danish and Norwegian are nearly identical, the questionnaire was translated into Norwegian by adapting a Danish translation (Christina Dideriksen, personal communication) conducted by a professional translator. To check the translation for quality, it was further forward-back translated by a bilingual team and adjusted if needed.

Communicative Development Inventories. The MacArthur-Bates Communicative Development Inventories (CDI; Fenson et al., 2007) are parent-report measures of early language comprehension and production, administered in the form of vocabulary checklists. We asked parents to check the words that their child understands and produces (for infants between 8 and 18 months old) or produces (for toddlers between 18 and 36 months old) and we collected CDI scores to measure children’s vocabulary sizes. We used adaptations of the CDIs for the relevant languages, namely the Norwegian CDIs (Simonsen et al., 2013), the Russian CDIs (Vershinina et al., 2011), and the Oxford CDI (Hamilton et al., 2000). Given that the CDIs are only appropriate to be used with infants between 8 and 36 months of age, from our initial sample of 3333 parents, we collected CDI comprehension scores for 1217 children (n = 315 from Norway, n = 559 from the U.K., and n = 343 from Russia), and CDI production scores for 2810 children (n = 795 from Norway, n = 559 from the U.K., and n = 1456 from Russia).

Following the procedure described in Kartushina et al. (2022), the following inclusion criteria were used to recruit participants: (a) monolingual children, defined as having a minimum of 90% exposure to their native language, according to caregivers’ reports, (b) full-term babies, defined as born at 37 weeks of gestation or later, (c) no diagnosed developmental disorder, and (d) no hearing/vision impairment. Therefore, participants who did not meet these requirements were not included in the study. We excluded individuals when we were unable to match participant ID and/or date of birth across questionnaires. Moreover, given that some parents filled in the questionnaire even though their children were older than 36 months, we excluded from the vocabulary analyses the children that did not fall into the age range (n = 301). Given that the CDIs contain different number of items across languages, in order to be able to compare infants’ vocabulary within and between languages, as well as across ages, the raw CDI scores were transformed into daily percentiles using normative data available at wordbank.stanford.edu (Frank et al., 2017) and following the procedure described in Kartushina et al. (2022). They were then divided by 100 such that values were bound between 0 and 1 to satisfy the assumptions of beta regression models.

To examine the psychometric properties of the EPAQ across the three languages and countries (U.K., Norway, Russia), a series of measurement invariance tests were conducted on the original three-factor solution proposed by Hembacher and Frank (2020). Configural, metric, and scalar invariance were assessed using the comparative fit index (CFI), root mean square error of approximation (RMSEA), and standardized root mean square residual (SRMR), as these indexes are less affected by sample size. Chi-square tests are known to be of little interpretative value in larger samples because, as sample size increases, the model’s statistical power becomes large, and the null hypothesis is more likely to be rejected, even with minimal inaccuracies (Zheng & Bentler, 2024).

As measurement invariance was not achieved, separate exploratory factor analyses (EFA) with oblique rotation were performed for each country to identify the underlying factor structure. Noninvariance is indicated by change of ≥ -.010 in CFI along with either a change of ≥ .015 in RMSEA or a change of ≥ .030 SRMR (Chen, 2007). The cutoff values needed to conclude that there is a relatively good fit of the model are: a value ≥ .95 for CFI, a value ≤.08 for SRMR, and a value ≤ to .06 for RMSEA (Hu & Bentler, 1999).

For the U.K. and Norway samples, which showed a relatively similar factor structure, the fit and measurement invariance of a two-factor solution based on the EFA results were estimated. Internal consistency reliability of the resulting scales was assessed using Cronbach’s alpha and mean scores were calculated. For Russia, three mean scores were created based on the unique three-factor solution that emerged. Please note that these mean scores were calculated based on the EFA and CFA results, not on the original EPAQ structure, which was not supported.

To investigate the relationship between parental attitudes and beliefs and children’s vocabulary development, two sets of models were tested. The first set (Models 1a and 1b) examined whether the factors identified in Norway and the U.K., and their interactions with child age, predicted children’s vocabulary scores in comprehension and production. The second set (Models 2a and 2b) tested the same relationships using the factors identified in the Russian sample. Full-null model comparisons were conducted to assess the significance of the combined predictors and to control for type-I errors. Estimates from the full models were inspected to identify significant associations between specific factors, their interactions with age, and children’s vocabulary scores, while controlling for gender and parental education. What follows is a detailed description of the analytic procedures.

We first tested for the measurement invariance (e.g., Schmitt & Kuljanin, 2008; Vandenberg & Lance, 2000) of the three-factor solution proposed by Hembacher and Frank (2020) using structural equation modeling with lavaan in R (Rosseel, 2012). Here, we compared a model that assumed the same three-factor structure for all three countries (i.e., testing for configural invariance, the lowest level of invariance; e.g., Meredith, 1993; Vandenberg & Lance, 2000), to a model that assumed the factor loadings to be the same (i.e., as in metric invariance; e.g., Horn & Mcardle, 1992; Vandenberg & Lance, 2000), and a model in which the item intercepts are assumed to be the same in addition (as in scalar invariance, the highest level of invariance; e.g., Vandenberg & Lance, 2000). To compare covariation between countries, metric invariance needs to be met (Jiang et al., 2017). To validly compare the means of the variables, scalar invariance needs to be met. Provided that the scale does not achieve measurement invariance, we planned to test alternative factor solutions based on item inspections, modification indices and insights from exploratory factor analyses in the different countries.

To estimate the extent to which the scales that were established in this step predicted the vocabulary scores (percentiles) in comprehension and production, we adopted a full-null comparison framework to avoid “cryptic multiple testing” (i.e., interpreting results from different tests as if they arose from a single test) (Forstmeier & Schielzeth, 2011, p. 47). The null model contained maternal education and gender of the child as control variables. Both of them were z-transformed with the scale function in R (R Core Team, 2022) to ease model convergence. In addition to the control variables, the full models included the EPAQ factors identified in each country and their interaction with the child’s age (in days), given that the effect of parents’ behaviors may increase or decrease with the child’s age. When the full-null comparison was significant, we provided inference for individual effects by dropping them one at a time, using the drop1 function in R. Given diverging structures of the factor solutions that emerged in the previous steps between Norway and the UK, on the one hand, and Russia on the other (see Results), we decided to run the analyses for the Norwegian and the U.K. data conjointly, while the Russian data was analyzed separately.

The first set of the pre-registered models evaluated whether the EPAQ factors identified in Norway and the U.K. and their interaction with the child’s age (independent variables) predicted vocabulary scores (dependent variable) in comprehension (Model 1a) and in production (Model 1b). In both models, the correlation between the random slope and the intercept was removed to allow model convergence. To account for differences in the effect of SES across both countries, we added the following random structure to the model (1+ z.edu|country), where z.edu is the z-transformed maternal education level).

A second set of models aimed to evaluate whether the EPAQ factors identified in Russia and their interaction with the child’s age (independent variables) predicted vocabulary scores (dependent variable) in comprehension (model 2a) and in production (model 2b).

The results from the measurement invariance test of the original three-factor solution are presented in Table 2. As presented, the CFI and the SRMR partly indicated an unacceptable fit already at the configural level, whereas the RMSEA indicated a relatively acceptable fit of this model. Importantly, absolute CFI and SRMR when testing for higher levels of invariance exceeded the thresholds for the metric invariance test (Chen, 2007). The absolute CFI, RMSEA as well as SRMR also exceeded the threshold for the scalar model. Thus, measurement invariance was not achieved at neither the metric nor the scalar level; in addition, the configural model showed an unsatisfactory fit to the data.

The inspection of the modification indices and consequent changes to the model did not sufficiently improve the model fit. The low CFI observed at the configural level suggested that the baseline 3-factor model did not adequately represent the factor structure across the samples. As reverse-coded items may cluster together and impair model fit, we also tested a factor solution without the reversed items. However, this fit was still unacceptable (see Table S1 in the Supplementary Materials).

Thus, we ran separate exploratory factor analyses for each country. Based on these results, it became clear that the factor solution was entirely different in Russia from that in the U.K. and Norway and that the original three-factor solution would not replicate in any country. Table S4 in the Supplementary Materials includes the factor loadings obtained in the original U.S. sample by Hembacher & Frank (2020) and the factor loadings obtained in the present analyses. The factor loadings in the U.S. sample reveal that the factor structure of the original EPAQ scale already had substantial issues. Specifically, the scale includes 11 items with cross-loadings (eight cross-loadings between AA and EL and three cross-loadings between EL and RR). We present the results from the U.K. and Norway first, as these indicated a relatively similar three-factor solution, albeit differing from the solution proposed by Hembacher and Frank (2020), see Table 3.

As displayed, in the U.K. and Norway, six items without cross-loadings (+/- .32) loaded on the same first factor that essentially represents the original Rules and Respect. These are: “It is very important that children learn to respect adults, such as parents and teachers”, “it is okay if children see adults as equals rather than viewing them with respect*”, “It is very important that there are consequences when a child breaks a rule, big or small”, “It is okay if young children boss around their caregivers*”, “It is very important for young children to do as they are told, for example, waiting when they are told to wait”, and “Children should be grateful to their parents”.

Six items loaded without cross-loadings on the second factor that essentially represents the original Affection and Attachment factor. These are: “Children who receive too much attention from their parents become spoiled.*”, “Too much affection, as hugging and kissing, can make a child weak*”, “Parents should pay attention to what their child likes and dislikes”, “Children should be comforted when they are scared or unhappy.”, “Children and parents do not need to feel emotionally close as long as children are kept safe.*”, and “A child who has close bonds with his or her parents will have better relationships later on in life.”. Two items that dealt with preparing children for school, that were included in the original Early Learning factor, loaded on the third factor referred to as School Preparation. These are: “Parents can prepare young children to succeed in school by teaching them things, such as shapes and numbers.”, and “Children don’t need to learn about numbers and math until they go to school”.

Thus, we estimated the fit and measurement invariance of this two-factor solution in the U.K. and Norway. A third factor was not included, as a measurement invariance test requires at least three items per factor. This two-factor solution showed acceptable metric invariance on all indices (see Table 4). However, the changes in the fit indices did not support scalar invariance. The reliability of the Rules and Respect scale was acceptable (U.K.: α = .74; Norway: α = .70). However, the Affect and Attachment scale had unacceptable reliability that could not be improved by omitting items with low inter-item correlations (U.K.: α = .56; Norway: α = .56). The correlation between the two school items was moderate (U.K.: r = .48, p < .001; Norway: r = .54, p < .001). As a last attempt to achieve a common factor solution, we conducted an additional invariance test of the new EFA-based 2-factor solution found in the U.K. and Norway, including the Russian sample in the analysis. However, including this sample drastically reduces model fit, as shown in Table S2, again indicating that it is unfeasible to detect the same factor solution across countries. Thus, only for the U.K. and Norway, we each created two mean scores measuring rules and respect (U.K.: α = .74; Norway: α = .70) and affection and attachment (U.K.: α = .56; Norway: α = .56).

In Russia, an entirely different 3-factor solution emerged. The first factor represented a mix of six items without cross-loadings from each of the three original dimensions and could be best described as Adult Authority (α = .70). This factor described parenting beliefs mainly valued in a traditional environment, based on respect, learning, and emotional control. Next, a total of eight items without cross-loadings, again representing a mix of the original three subscales, loaded on the second factor. This factor could be best described as Communicative and Emotional Detachment (α = .77) and characterizes parenting attitudes based on the idea of a natural learning process and a risk of spoiling the child with excessive attention. Finally, without cross-loadings, two items loaded positively and one item negatively on the third factor that could be described as Confrontation Between Child and Adult, but the resulting scale had unacceptable reliability (α = .32; the negative item was reversed before calculations). This third factor defines parenting beliefs that value the importance of a strict hierarchical structure and does not promote children’s independence.

Thus, for Russia, we created three mean scores measuring adult authority (α = .70), communicative and emotional detachment (α = .77), and confrontation between child and adult (α = .32).

Models 1a and 1b: Parental Attitudes Predicting Vocabulary Development in Norwegian and U.K. Children. The first set of models (1a and 1b) investigated whether the factors identified in Norway and the U.K. and their interaction with the child’s age predicted children’s vocabulary scores in comprehension and production. The full-null comparisons did not reveal any significant improvements, neither for comprehension (χ² = 4.33, p = .741) nor for production (χ² = 3.40, p = .846), suggesting that differences in parental beliefs and attitudes, as indexed by the three factors identified in Norway and the UK, did not explain children’s vocabulary sizes.

Models 2a and 2b: Parental Attitudes Predicting Vocabulary Development in Russian Children. The second set of models (2a and 2b) investigated whether the factors identified in Russia and their interaction with the child’s age predicted children’s vocabulary scores in comprehension and production. The full-null comparison was significant in production (χ² = 18.87, p = .009), but not in comprehension (χ² = 11.25, p = .128). The former indicated that combining the three factors (factor1_ru, factor2_ru, factor3_ru) and their interaction with z.age significantly improved the model fit for production. An inspection of the estimates in the full model for production (see Table 7) revealed that the interaction between the second factor identified in Russia, namely Communicative and Emotional Detachment, and age was significantly negatively associated with children’s vocabulary scores in production.

Children whose parents held parenting beliefs described by the factor Communicative and Emotional Detachment were reported to have lower expressive vocabulary size relative to the normative (age-matched) data, with younger children being affected more than older children (Figure S1). A gender effect was also observed (p < 0.001) in the control predictors, suggesting that girls were better than boys regarding speech production.

Previous research has reported that intuitive parenting theories predicted parents’ actual parenting behavior (Hembacher & Frank, 2020). The present study was designed (1) to cross-culturally assess the psychometric properties of the Early Parenting Attitudes Questionnaire (EPAQ) by Hembacher and Frank (2020) in Russia, Norway, and the U.K. and (2) to test pre-registered hypotheses regarding the relationship between parenting attitudes and beliefs and children’s vocabulary scores.

In line with previous findings in Russia with parents to older children (Bukhalenkova et al., 2021), an exploratory factor analysis of the parental responses on EPAQ highlighted an entirely different factor solution in Russia, Norway, and the U.K., as compared to the original three-factor solution found in the American sample (see table S4 in the Supplementary Materials). The original EPAQ scale already presents structural issues, especially regarding the second factor (i.e. Early Learning). In fact, several items that the Hembacher and Frank (2020) included in the EL subscale load strongly on the AA factor, and several AA and RR items load strongly on the EL factor. The factors obtained in Norway and the U.K. only partly replicated the original EPAQ structure and differed significantly from the ones that emerged in Russia. Therefore, two alternative factor solutions were proposed. Our interpretation relies on the factor descriptions of the original authors, and we adjusted them when the factor loadings we obtained suggested otherwise. However, it is important to note that the interpretation of factor solutions is a challenging process, given that it is based on culturally grounded interpretations of items that load together on one factor. Therefore, the factor labels could change depending on the way one interprets the item loadings. In the U.K. and Norway, the first two factors seemed to represent the original Rules and Respect and Affection and Attachment subscales. A third factor, different from the original one, was identified, namely School Preparation. In Russia, an entirely different solution emerged: each factor included items of the three original EPAQ subscales (i.e., Affection and Attachment, Early Learning, and Rules and Respect). Following a rigorous consideration of the obtained factors, it was determined that they could be referred to as Adult Authority, Communicative and Emotional Detachment, and Confrontation Between Child and Adult. Adult Authority describes a parenting attitude mostly valued in a traditional environment based on respect, learning, and emotional control. Communicative and Emotional Detachment characterizes a parenting attitude based on the idea of a natural learning process and a risk of spoiling the child with excessive attention. Confrontation Between Child and Adult defines parenting beliefs that value a strict hierarchical structure and do not promote children’s independence.

Given that linguistic differences often do not allow for direct translations, it is essential to use a translation approach that follows standard procedures. Language differences can be a source of invariance and potentially change the latent factor structure. However, we argue that, in this case, the lack of measurement invariance found in the analyses may also be attributed to significant cross-cultural differences in parenting rather than solely linguistic differences. Although some parenting behaviors are likely to be similar across cultures (i.e., physical caregiving), there are also considerable differences due to the influence of other factors (Lansford, 2022), such as the number of siblings in the family (Kramer & Hamilton, 2019), the expected behaviors from the parents (Lansford et al., 2018), and the type of cognitive stimulation provided (Bornstein et al., 2015). Therefore, these differences might indicate that parenting beliefs are not as generalizable as one might think and that they could be accountable for the lack of measurement invariance. However, we highlight once more that language and culture are highly related, and it is, therefore, very difficult to distinguish them in most contexts (with the exception of different cultures using the same language).

In the developmental literature, it is not unusual that scales developed in the West fail to replicate in other parts of the world. Ruchkin and colleagues (2007) tried to validate the Strengths and Difficulties Questionnaire (SDQ; Goodman, 1997) in a sample of Russian adolescents and found that the Russian version of the questionnaire had unsatisfactory psychometric properties. A study by Vu and colleagues (2019) highlighted the importance of considering cross-cultural differences when measuring parenting beliefs and attitudes about feeding. They found that, consistent with previous research (Liu et al., 2014), the original structure of the Child Feeding Questionnaire (CFQ; Birch et al., 2001), developed in the U.S., failed to capture cultural-specific beliefs of immigrant Chinese families even when they lived in Western contexts. Within the West, Gjersing and colleagues (2010) found that the original scale to measure staff attitudes towards Opioid Maintenance Treatment (OMT; Caplehorn et al., 1998), developed in Australia, failed to grasp essential concepts and beliefs in the Norwegian setting, while both Australian and Norwegian societies are considered as Western.

Contrary to our expectations, our study did not reveal any significant association between parental beliefs and vocabulary scores in Norway and the U.K., suggesting a lack of predictive validity for the scale. We expected that the factor referred to as School Preparation would associate positively with CDI scores, as parent involvement in school is usually associated with good academic achievements (e.g., Barnard, 2004; Zellman & Waterman, 1998). On the other hand, we expected that Rules and Respect would correlate with lower CDI scores, as children living in this environment could be hesitant to ask questions and try new things as they might be unsure of their parents’ approval. Moreover, given that the majority of the items that loaded on Affection and Attachment are related to the child’s safety and emotional well-being, we expected that this factor would correlate positively with vocabulary scores. In fact, a child living in a safe environment and developing a secure attachment feels comfortable speaking and interacting with others (Bowlby, 1969/1982). Considering the lack of measurement invariance found in the analysis and given that no significant associations have been identified between the EPAQ and CDI scores, it might be reasonable to assume that the implicit parenting theories measured by the EPAQ do not actually predict parenting behaviors or outcomes (e.g., speech comprehension and production) in Norway and the U.K.

In Russia, considering that the factor Adult Authority describes a parenting attitude based on respect, learning, and emotional control, we expected to find a positive correlation with the vocabulary scores. We assumed that Russian parents with these parenting beliefs value independence while still holding their kids to high standards, allowing them to grow confidently and maximize learning opportunities (Cerino, 2021). However, no significant association was found between Adult Authority and vocabulary scores.

Our study found a significant negative association between the factor Communicative and Emotional Detachment and vocabulary scores in Russia, as predicted. According to McCafferty (2002), Meins (1997), and Karass et al. (2003), these parents do not encourage independence and are not sensitive-responsive, a trait that attenuates the development of language and communication. Such parents share the belief that excessive attention is harmful to the child and that if a child does not speak, then they do not understand speech either. Consequently, they tend to limit their engagement in activities that prior studies found to promote language development, such as shared book reading (Kartushina et al., 2022; Shahaeian et al., 2018), speaking (Rowe, 2018), and playing (Hirsh-Pasek, 2009). Given that language develops in social interactions (Kuhl, 2007), some caregiving behaviors have been found to be responsible for linguistic development (Bruce et al., 2022). A study by Vallotton and colleagues (2017) analyzed sensitivity and cognitive stimulation to assess their effect on vocabulary development over time. Sensitivity refers to the warm and timely reactions that parents have when the child interprets and responds correctly to cues (Shin et al., 2008). In contrast, cognitive stimulation refers to parents’ efforts to engage in activities that promote cognitive development (Martin et al., 2007). They found that both aspects critically influence child vocabulary development in the first three years of life; however, sensitivity has a more significant impact during early development, while stimulation becomes more important later. This also supports our finding that younger children are affected more by their parents’ emotional coldness than older children. Nevertheless, despite the single association between one EPAQ factor and the CDI observed in Russia, the predictive validity of the scale was generally not substantiated. Considering the potential for false positives given the extensive number of tests conducted, we are cautious about placing significant weight on the one positive finding in Russia. A scale’s utility hinges not only on its cross-cultural factorial validity but also on its capacity to predict meaningful outcomes across diverse cultures and contexts. The EPAQ currently appears ill-suited for elucidating a substantial portion of language development in children. This limitation may be attributable to the scale’s deficient psychometric properties that we observed in this study or to its potential measurement of concepts of scant relevance to the outcome of interest.

Furthermore, a gender effect in expressive vocabulary was observed in the Russian sample, suggesting, in line with previous research, that girls perform better than boys when it comes to language production (e.g., Adani & Cepanec, 2019; Lange et al., 2016; Mccarthy, 1953).

The results of the present study should be interpreted in light of some limitations. First, although the sample was gender-balanced, it mainly consisted of parents with a medium-to-high level of education. Future studies should try to administer the EPAQ to a sample with a larger proportion of parents with a low level of education. Second, our sample included two Western countries (i.e., Norway and the U.K.), and one Eastern country (i.e., Russia). In future research, additional countries could be included in the sample in order to further test the generalizability of the results. Third, the type of data collected in the present study does not allow for causal conclusions, although correlational data may inform causal conclusions in some specific cases (see Pearl, 2000; Rohrer, 2018).

On the basis of the results of the present study, we conclude that the EPAQ by Hembacher and Frank (2020), designed to measure intuitive theories of parenting, should be used with caution in its current form to measure parental attitudes and beliefs in parents living in other contexts than the North American. These findings raise the question of whether implicit parenting theories are similar across cultures to an extent that allows for their assessment and comparison with psychometric scales. Likely, scales such as the EPAQ need to be adapted in order to comply with the cultural and societal norms of the respective countries and be generalizable across contexts. Moreover, the EPAQ’s underlying structure revealed in the Norwegian and the UK samples appears to have limited predictive validity in terms of infants’ language production and comprehension.

Admittedly, the EPAQ was developed in a North American context without the ambition to measure a cross-culturally valid phenomenon. As we find some evidence that parts of its structure generalize to contexts with similar Western languages (English, Norwegian) and cultures (U.K., Norway), it may be further developed to capture the same structure of parenting attitudes in this region of the world. This would require an in-depth analysis of parenting beliefs across the respective cultures and a careful testing of items. However, the structure of parenting attitudes seems too different in our third country, Russia. Thus, for such contexts, different culture-sensitive measures need to be developed.

More research is also needed to test for the consequences of implicit parenting theories for developmental markers. Surprisingly, the subscales showed little of a relationship with language abilities, as measured by the CDI. Therefore, future research is needed to establish whether the EPAQ predicts other developmental outcomes in meaningful ways. Alternatively, parental attitudes at these early ages do not relate to infants’ very early language skills, which, however, would be at odds with previous research suggesting that early parent-child interactions promote language and communicative skills (e.g., Cartmill et al., 2013; Hirsh-Pasek, 2009; Kartushina et al., 2022; Vallotton et al., 2017). Yet, it is possible that the reason for why the EPAQ has low predictive power is that it only translates into specific behavioral patterns to limited degrees.

Deepening the knowledge of the implicit theories of parenting can provide significant insights into how and to what extent the environment exerts an influence on several aspects of children’s development. This line of research may be useful to create effective interventions to support families in understanding and implementing the best strategies to promote healthy development for their offspring.

The authors certify that they have no affiliations with or involvement in any organization or entity with any financial interest, or non-financial interest in the subject matter or materials discussed in this manuscript.

Conceptualization, NK, JM

Preregistration, SA, NK, JM

Data Collection, MG, NK, JM

Data Processing, SA, NK, JM, JRK

Data Interpretation, SA, MG, NK, JM, MR, JRK

Writing, SA, MG, NK, JM, JRK

Review & Editing, SA, NK, JM, MR, JRK

Funding Acquisition, NK, MG, MR, JM