Participants were eligible for the study if they were mothers of a child within the middle childhood age range (6–10 years old). Because most existing research on intensive parenting has been done among parents with adolescent children (Ryan et al., 2024), there is limited information about the parenting of younger children. Using mothers as informants maximises the ecological validity of IPB Scale target behaviours, such as daily engagement with the children, monitoring them and managing their activity. Furthermore, the South African sample, with its diversity of cultural norms, schooling quality and safety concerns, would have differential effects on parenting behaviours, thus addressing an area specifically lacking in documented research.

Mothers were recruited online through social media platforms by posting the research advertisement as a flyer on open Facebook parenting groups. The flyer contained information on the aims and nature of the study and the link for the survey, and voluntary participants were requested to complete the online (Qualtrics) survey anonymously. The invitation to participate in the research remained open for 2 months until 507 mothers completed the survey. It was thus a self-selected volunteering sample of mothers.

A cross-sectional online survey design was used to determine the degree of mothers’ IPB by means of an adapted IPB Scale. The aim was to evaluate the functioning of the scale, among others, by exploring its factor structure. This quantitative study assumes a critical realist stance, thus assuming latent constructs to be real but also that they are epistemically dependent on the quantitative approach used, thus rendering our inferences about them corrigible (Bhaskar, 2008; Boost et al., 2022).

Mothers completed the 69-item IPB Scale, targeting mothers of children in middle childhood (age 6–10 years), developed in Poland (Lubiewska et al., 2024). Mothers rated the extent to which they practise certain parenting behaviours with their child between the ages of 6 years and 10 years on a 7-item Likert response scale ranging from (1) never to (7) always or (1) definitely no to (7) definitely yes. An example item is ‘I encourage my child to make his/her own decisions and take responsibility for their choices’. In addition to the IPB Scale, the questionnaire contained demographic mother-related information such as age, marital status, level of education, socio-economic status, ethnicity, vernacular and location where she stays (rural or urban) and child-related factors such as health, schooling, learning difficulties and extracurricular activities to contextualise the survey in the South African context.

The IPB Scale was adapted for use in South Africa to accommodate linguistic, cultural and contextual nuances, ensuring relevance across diverse communities. Specifically, (1) language was simplified and adapted to terminology commonly used across South Africa (e.g. ‘secondary’ changed to ‘high’ school; ‘extracurricular’ changed to ‘after school’) and (2) socio-economic inequalities were accounted for by adapting items measuring provision of material items (e.g. ‘if possible’ was added to ‘I give my child more than what he/she asks for’). To ensure linguistic and cultural relevance, a cognitive debriefing interview was conducted, which is considered good practice in cross-cultural adaptation of instruments (Lee, 2014). The interview was conducted with a multilingual black African mother familiar with three cultural groups, four languages and rural upbringing, thus able to reflect on diverse cultural and linguistic contexts. The interview identified wording complexity (e.g. simplifying terms such as ‘extracurricular’ to ‘after school’) and highlighted cultural nuances regarding autonomy. As a result, two items were added to account for cultural variance related to child autonomy, differentiating between maternal beliefs (‘I make decisions for my child, as children 6–10 years old cannot make decisions’ indicated that children in this age range are not cognitively capable of making decisions; therefore, mothers should make them on their behalf) and cultural norms (‘Children of 6–10 years old don’t have a say’, as in some cultures in South Africa, parents make decisions on behalf of the child). Although only one cognitive interview was undertaken, this step provided insight into item interpretation and informed the adaptation process. Finally, pretesting was done with two participants for ease of use on a mobile device (the device most likely to be used). This was done to ensure that the survey was readable, clear and easy to navigate on a mobile device. Their feedback was implemented with final minor adjustments to ensure understanding across a range of literacy levels. The minor adjustments included removing the use of a double negative in a sentence, which made the question confusing, and the orientation of the Likert scale responses (vertical as opposed to horizontal) to ensure mobile-friendly interaction.

Prior to participation, as part of the informed consent procedure, the mothers were informed about the study aims, the general content of the items and the possibility to withdraw from the survey without negative consequences. Before they could complete the survey, they had to provide consent. Informed consent was therefore obtained through an electronic consent form that participants were required to read and agree to by actively selecting to consent before accessing the survey. If they chose not to consent, they were unable to proceed to the survey. After completing the survey, participants received a brochure (developed by the head of the Parenting and Attachment across Cultures Lab and thus appropriate for the target population) with age-appropriate parenting information, which noted that parenting practices may vary across families and children. At the end of the study, mothers were also invited to voluntarily provide their email addresses, collected separately from the survey data to preserve anonymity and ensure no identifying information could be linked to participants’ responses, for entry in a raffle for vouchers that could be used at a well-known online store, as a thank you for participation. Participation in the raffle was entirely voluntary and not a requirement for completing the survey. Provision was made for participants to contact the research team members if they felt distressed as a result of participating in the study. To maintain confidentiality, the data were stored securely on password-protected devices and were only available to the researchers.

An exploratory factor analysis (EFA) was done to determine the factor structure of the instrument. The requirements for an EFA are considered and indicated below, starting with the sample size considerations.

Empirical evidence shows that rules of thumb for sample size discount the complexity of considerations for EFA (Henson & Roberts, 2006). Thus, sample size does not straightforwardly determine the adequacy of a factor structure, which makes a priori decisions regarding sample sizes difficult (Henson & Roberts, 2006). Small sample sizes can be tolerated when communalities are high (> 0.60) and there are at least four or more items per factor (MacCallum et al., 1999). With low communalities (< 0.50), a larger sample size (> 100) and more indicators per factor (six to seven) are important, along with a smaller number of factors. When there are only three to four indicators per factor and communalities are low, the sample size should be at least 300 or more (MacCallum et al., 1999, p. 96). Missing values were deleted listwise, resulting in a final sample of N = 382 with complete responses. This yielded a variable-to-sample ratio of 1:18. The data were checked for distributional properties, indicating normality for each variable and/or item by Kolmogorov–Smirnov with Lilliefors adjustment (p > 0.05). Extreme skewness (> 2) and kurtosis (> 7) were identified (Fabrigar et al., 1999). The variable distribution was also visually inspected.

Bartlett’s test of sphericity and the Kaiser–Meyer–Olkin (KMO) test were used to determine the suitability of the dataset for EFA. Kaiser–Meyer–Olkin values above 0.60 support the use of EFA, and Bartlett’s test should be significant (Field, 2018). The software used for analysis was Statistical Package for Social Sciences (SPSS) version 30 (IBM Corp., 2024) and R (R Core Team, 2024). The items used in the EFA had seven categories and sufficient distribution across these to be regarded as continuous (Tabachnick & Fidell, 2013). The Pearson correlation matrix was used for the EFA.

Principal axis factoring (PAF) was used as the factor model, with maximum likelihood (ML) as the extraction method. Communalities were determined with ML, both before and after extraction. Oblique rotation, specifically Promax, was used because the factors were expected to correlate. Promax begins with a varimax rotation to force a simple structure and then follows with an oblique rotation (Howard, 2016).

The number of factors to extract was based on the following criteria (with the suggested number of factors in brackets): eigenvalue > 1 (15 factors) (Kaiser, 1960), Velicer’s minimum average partial (MAP) test (seven factors) (O’Connor, 2000; Velicer, 1976), very simple structure (VSS) or simple MAP (two factors), scree plot (four or seven factors) (Cattell, 1966) and parallel analysis (PA) (eight factors) (Horn, 1965). Kaiser’s criterion and the scree plot were not considered because of low accuracy (Taherdoost et al., 2014). The MAP suggestion of the seven factors was taken as a guide, along with the interpretability of the factor structure, which showed that extracting more than seven factors resulted in too few items per factor.

The factor interpretation strategy was based on at least three items loading above 0.30 on the target factor (Watkins, 2018), although Henson and Roberts (2006) recommend at least two items. Cross-loadings should be below 0.30 (Howard, 2016). The pattern matrix is reported with values of 0.25 and higher. Communalities after extraction are reported, as well as inter-factor correlations. Given the fact that we do not believe the sample is representative of the population of mothers in South Africa and the contingent nature of the realised sample, it is possible that items with low communalities and/or low loadings could function better or even worse in subsequent surveys in South Africa. We thus assumed a relatively conservative item selection strategy, and in the face of a lack of additional evidence to delete a few poorer functioning times, they were retained, as can be seen in the results section below. Therefore, reliability estimates for the factors – in this case, Cronbach’s alpha (α) – were reported for the retained factors after extraction but also after items were shifted or discarded.

A second-order factor was conducted with the factor scores from the first EFA using the same criteria as the first-order analysis. Previous research identified two bipolar constructs, which were investigated using the current dataset. In this instance, Kaiser’s criterion and the scree plot concurred with an extremely clear gap between the second and third eigenvalues.

Ethical clearance to conduct this study was obtained from the Ethics Committee of the Faculty of Humanities, University of Pretoria (No. HUM010/0124).

The sample demographic characteristics are set out in Table 1. One of the most striking findings in Table 1 is the substantial amount of missing data regarding employment type, with 52.6% of responses classified as missing for the part-time versus full-time employment status. This figure represents the highest missing data rate within the dataset. Furthermore, the income distribution among respondents highlights notable economic challenges. A considerable portion – 63.3% of participants – reported earning less than R20 000.00 per month, encompassing the first three income brackets. Conversely, only 2.4% of respondents reported earnings exceeding R70 000.00, indicating a highly skewed income distribution that favours lower earnings. These income data suggest that the sample predominantly represents a lower-to-middle-income population.

Patterns regarding relationship status also emerged from the analysis. The category ‘unmarried but in a relationship’ was the largest segment, accounting for 38.2% of respondents, surpassing the married category at 32.5%. When combining all unmarried individuals, they constitute 50.8% of the sample, significantly outnumbering the married individuals. The data reveal low numbers of divorcees (2.1%).

In terms of demographic composition, the sample is predominantly composed of black African respondents, who make up 68.8% of the population. Other ethnic groups are represented in much smaller proportions: white respondents account for 9.2%, mixed race respondents account for 3.7% and Indian respondents account for 1.8%, which roughly reflects the South African population (Statistics SA, 2024). The analysis also indicates a high level of language diversity within the sample, with 12 different languages represented; however, isiZulu is the most spoken language, dominating at 18.1%.

The data further uncover significant employment challenges within the sample, as evidenced by a notably high unemployment rate of 37.2%, which reflects the current high national unemployment rate. Among those who are employed, full-time work is significantly more prevalent, comprising 35.9% of the workforce, compared to only 11.5% engaged in part-time employment. Additionally, a consistent pattern of missing data across most demographic variables was observed, with rates ranging from 14.4% to 16.5%. This indicates that a specific group of participants did not respond as expected – potentially because of survey fatigue, as the demographic section appeared at the end of the survey and may have been too long for some respondents.

Overall, the analysis of the data indicates that the sample reflects a South African population grappling with economic challenges, high unemployment rates and significant data quality issues, particularly concerning the classification of employment status.

Only seven items (9%) were markedly skewed, while four items (5.6%) had a kurtosis above |7|. The four items were also skewed > |2|, namely, IP2b.1, IP2b.4, IP2c.1 and IP2c.4. These items were flagged. Excluding these outliers, the average |skewness| was 0.90 (standard deviation [s.d.] = 0.56) and |kurtosis| = 1.26 (s.d. = 1.2).

Under the adaptive IPBs, the first-order factors loaded positively. Engagement reflects maternal involvement, support and interest in her child’s life. In South Africa, where the quality of early education and childcare varies (Hatch & Posel, 2018), maternal involvement may enable optimal child development, because active parental engagement is known to promote child development (Baker et al., 2018). Autonomy support fosters child independence by encouraging autonomous problem-solving and decision-making and the development of their own interests. Research suggests that autonomy-supportive involvement from parents is associated with positive child outcomes in various contexts (Lerner et al., 2022), including Africa (Marbell-Pierre et al., 2019). Behavioural control relates to regulating the child’s behaviour through monitoring, rule setting and guidance. This factor included four items related to academic involvement. This contrasts with the international data (Lubiewska et al., 2025), where academic engagement formed a separate factor within the child undermining the second-order dimension. Consolidating academic involvement into behavioural control may be a context-specific feature. Educational and socio-economic inequality is characteristic in South Africa, and maternal involvement may therefore be an adaptive strategy, as research indicates that parental academic involvement is an important predictor of academic achievement and upward social mobility in contexts of economic inequality (Zhang et al., 2020). Maternal monitoring and guidance in this area may therefore compensate for systemic shortcomings and align with broader literature associating behavioural control with limit setting and academics and with adaptive parenting resulting in positive child outcomes (Doan et al., 2017). As such, what may be classified as undermining in contexts with more resources may be considered supportive in South Africa.

Given South Africa’s historical, social and cultural context, characterised by high socio-economic inequality and crime rates and low literacy and education levels, maternal engagement, autonomy support and behavioural control (rule setting, academic involvement, monitoring and guidance) are therefore considered adaptive IPBs. Extant literature suggests they encourage optimal child development, appropriate mother–child attachment and balanced interaction by combining responsiveness and support with boundaries and expectations within this context (Baker et al., 2018; Lerner et al., 2022; Yan et al., 2017).

Under the overinvolved IPBs, the remaining first-order factors loaded negatively. Autonomy restriction may potentially inhibit child development through maternal overinvolvement in daily tasks and infantilisation. This may prevent children from acquiring developmentally appropriate skills and contribute to maternal burnout by requiring mothers to perform tasks instead of the child (Shah et al., 2023). This factor contained the two items not in the original IPB Scale to differentiate between maternal beliefs and cultural norms regarding children’s decision-making ability, both relating to autonomy restriction, but for different reasons. Proactive protection is aimed at safeguarding the child from harmful circumstances or criticisms, aligning with ‘lawnmower parenting’, where parents remove obstacles out of their children’s way (Locke et al., 2012). While potentially well-intended or contextually necessary (e.g. high inequality and/or crime), this may undermine the development of, among others, resilience and coping strategies (Flamant et al., 2022) and requires intensive maternal energy to constantly seek out and prevent potential negative experiences. Talent stimulation involves maternal focus on child stimulation through extramural activities and achievement pressure. If excessive, this can leave little time for independent exploration and relaxation (Tany & Khanam, 2022) and requires significant maternal time and financial investment. This aligns with the concept of ‘tiger parenting’, a strict and developmentally focused parenting style (Xie & Li, 2018). Contextually (e.g. low socio-economic status), higher levels of talent stimulation may enable child and/or family achievement and thus upward social mobility (Zhang et al., 2020). One item (‘I encourage my child to discuss any academic problems with his/her teachers’) cross-loaded weakly on autonomy support and should be flagged for future review, as it could reflect either teaching help-seeking skills (autonomy support) or pushing academic development (talent stimulation). Finally, child-centredness prioritises the child’s desires and requirements over those of the mother and/or family. This requires substantial maternal time, energy and finances, potentially encouraging materialism and decreasing the emotional well-being of children (Denegri et al., 2022; Trzcińska et al., 2024). However, research on poverty and well-being (material and subjective) in children suggests mothers may sacrifice their own needs to protect children from feelings of deprivation or hardship (Gross-Manos & Bradshaw, 2022).

Prior literature generally associates high levels of autonomy restriction, proactive protection, talent stimulation and child-centredness with detrimental outcomes, potentially undermining child development and placing excessive demands on maternal resources (e.g. Flamant et al., 2022; Locke et al., 2012; Shah et al., 2023; Tany & Khanam, 2022). However, contextual factors are important: in settings like low socio-economic status or low literacy, greater maternal involvement may not be maladaptive, potentially ensuring child and/or family safety, well-being and achievement.

The identified multidimensional structure aligns with prior studies, suggesting intensive parenting is a complex construct where not all aspects are uniformly adaptive or maladaptive. The factors child-centredness and talent stimulation were identified using other measures in cross-national studies by Gauthier et al. (2021) and Liss et al. (2013). The current structure also incorporates the resource-intensive component originally identified by Hays (1996). Specifically, high levels of proactive protection, child-centredness (including providing material goods) and talent stimulation require significant cognitive, emotional, financial and temporal investments. Given the cultural variance in South Africa and cross-national findings, child-centredness and stimulation appear to be core features of intensive parenting more broadly.

Compared with the IPB international results (Lubiewska et al., 2025), the South African data share not only similarities (e.g. first-order factors: engagement, autonomy support, talent stimulation [termed facilitation internationally], proactive protection and autonomy restriction) but also differences in factor structure. This is to be expected given the heterogeneous nature of the South African population. Child-centredness loaded on two separate factors in the international data (prioritising and material indulgence). This may reflect the specific meaning that providing material goods may have in a context of economic constraint, and in that, instead of indulgence, it may represent a form of protection against feelings of deprivation. This is supported by research on poverty and child well-being (Gross-Manos & Bradshaw, 2022) and the association between material resources and lower levels of child well-being across various countries (including Africa) (Main et al., 2019; Sarriera et al., 2015). However, further qualitative research in this regard is required. Two items on this factor were not retained in the international structure (doing things for the child rather than seeing the child worried and worrying too much about the child). The first indicates prioritising the child’s emotional well-being and is therefore appropriately loaded. The maternal worry item cross-loaded with similar values on three factors. It is thus not unique to this factor and was removed locally as well.

An item on prioritising the child’s interests over the mother’s loaded on engagement in the South African structure (but on prioritising internationally), potentially because of literacy levels and complex wording – the item was long and consisted of two parts: ‘I ensure that my child can pursue their interests; developing my own interests is less important to me’. It is possible that most mothers did not read the full question and answered based on the first half. Removing the second half is suggested, as it did not cross-load on any other factor.

Finally, behavioural control formed a single factor locally (adaptive dimension), whereas the international structure separates these items into academic engagement (undermining dimension) and rule setting (supportive dimension). As previously discussed, a contextual explanation may be that in South Africa (high educational and socio-economic inequality), active academic involvement and rule setting may be adaptive strategies supporting child development and potentially enabling upward social mobility of the family (Zhang et al., 2020). Additionally, in sub-Saharan Africa, parental involvement in children’s education is considered essential for development (Kambona, 2025). This factor also included two monitoring items (monitoring who the child spends time with and what happens at school) that were excluded internationally, but deemed adaptively important locally because of high crime and violence in the local context.

To synthesise these structural similarities and differences with the international IPB structure and prior conceptualisations of intensive parenting, a visual overview is provided in Figure 1. This mind map illustrates the alignment of the South African first-order factors within the identified adaptive and overinvolved second-order dimensions, while highlighting key divergences from the international IPB factor structure and core components identified in prior literature (e.g. Gauthier et al., 2021; Hays, 1996; Liss et al., 2013; Lubiewska et al., 2025).

This study is not without limitations. Firstly, despite the linguistic variance of South Africa, the language used in the measure was English. While the language of business and education in South Africa is predominantly English, and the measure was adapted for potential low literacy of respondents, the absence of translated versions may have influenced participants’ understanding of certain items. Additionally, only one cognitive debriefing interview was conducted, and while the participant was multilingual, additional interviews would have been beneficial. Future studies should therefore include multiple cognitive interviews to further strengthen the linguistic and cultural equivalence of the measure. Secondly, the sample was relatively small and not representative of all cultural groups, and results may therefore not capture intensive parenting across all cultural groups in South Africa. As most participants were low-to-middle-income mothers, the results may not reflect parenting behaviours of higher income groups or fathers. Generalisations should therefore be made with caution. Future studies should include more mothers from different cultural groups in South Africa and fathers, to establish the applicability of the IPB across South Africa’s population.