Autonomy support in toddlerhood: Similarities and contrasts between mothers and fathers.

Autonomy Support and Scaffolding of Children’s Goal-Directed Activity

Developmental science has transformed the understanding of young children’s social and cognitive competencies, leading to a shift in the focus of research on early parental influences. For example, whereas parenting studies have long focused on general constructs such as sensitivity or authoritative discipline, interest in young children’s executive functions ( Hughes & Devine, 2017a ) has led to multidimensional models and a renewed interest in specific constructs, such as autonomy support. Studies involving samples of different ages and from different countries ( Bernier, Carlson, Deschênes, & Matte-Gagné, 2012 ; Bernier et al., 2010 ; Hughes & Devine, 2017b ; Hughes & Ensor, 2009 ) have shown that both global measures of autonomy support and more specific measures of parental scaffolding predict gains in preschoolers’ executive functions, even when background parental measures are controlled. Recent evidence that early parental autonomy support predicts children’s later academic success ( Bernier, McMahon, & Perrier, 2017 ; Devine, Bignardi, & Hughes, 2016 ) has also highlighted the importance of early variation in parental autonomy support.

Do Mothers and Fathers Provide Similar Kinds and Levels of Autonomy Support?

Despite fathers’ having become increasingly involved in child care (e.g., Sullivan, Coltrane, McAnnally, & Altintas, 2009 ), they have, with recent exceptions ( Mills-Koonce et al., 2015 ; Nordahl, Zambrana, & Forgatch, 2016 ; Sethna et al., 2017 ), generally been overlooked in studies of individual differences in early parent–child interactions. Whereas early theorists (e.g., Belsky, Gilstrap, & Rovine, 1984 ) argued that differences in parental experience and/or context were likely to yield qualitative differences between mothers’ and fathers’ caregiving, recent theoretical reviews have called for gender-neutral models of parenting (e.g., Fagan, Day, Lamb, & Cabrera, 2014 ). Establishing the conceptual equivalence of maternal and paternal measures requires confirmatory factor analyses (CFAs) to test whether the model retains a good fit even when specific aspects (e.g., structure, variance, intercepts) are constrained to be equal across parent gender.

In turn, CFAs require larger samples than do those traditionally used in observational research. Indeed, just one previous observational study has tested measurement invariance in mothers’ and fathers’ caregiving. Applying CFAs of observational ratings of parental detachment, stimulation, positive regard, and animation at 6 and 24 months for 630 low-income North American families, Mills-Koonce et al. (2015) reported statistically equivalent overall ratings for mothers and fathers. That said, combining different aspects of parenting, different contexts, and different time points may have masked subtle nonequivalencies. Our first goal was thus to assess across-parent invariance for one specific measure (autonomy support) in one context and time point (an inset puzzle at 14 months).

Parental Autonomy Support: Individual and Dyadic Correlates

Framed by Belsky’s (1984) seminal account of the multifaceted determinants of parenting, a further goal was to examine autonomy support in relation to both parental personality and child gender and temperament. Meta-analytic results have indicated small but consistent links between personality and parenting ( Prinzie, Stams, Deković, Reijntjes, & Belsky, 2009 ) that appear similar in strength and nature for mothers and fathers ( de Haan, Prinzie, & Deković, 2009 ). However, widespread reliance on single-informant self-report methodologies limits these conclusions. Moreover, the few studies that included direct observations of parent–infant interactions (e.g., Clark, Kochanska, & Ready, 2000 ; Karreman, van Tuijl, van Aken, & Deković, 2008 ) focused on parental control and responsiveness or sensitivity rather than on autonomy support. To our knowledge, only one study involving mothers of preschoolers ( Neitzel & Stright, 2004 ) has reported a link between parental personality and observational ratings of autonomy support. Thus, our study adds to the field by involving both mothers and fathers of infants.

Turning to child characteristics, evidence has suggested a dosage effect in which children’s difficult behavior exerts a stronger influence on mothers’ than on fathers’ parenting (e.g., Elam, Chassin, Eisenberg, & Spinrad, 2017 ; Meunier, Roskam, & Browne, 2011 ). The gender composition of the parent–infant dyad has also appeared to be important. For example, in a study of parents and toddlers, Lovas (2005) reported a steady decline in parents’ emotional availability across mother–daughter, mother–son, father–daughter and father–son dyads. Likewise, whereas a meta-analysis of 126 observational studies of parents’ autonomy-support behaviors showed negligible effects of child gender ( Endendijk, Groeneveld, Bakermans-Kranenburg, & Mesman, 2016 ), moderation analyses revealed significant effects of both time and age. Specifically, studies from the 1970s and 1980s reported greater autonomy support for sons than daughters, but studies from the 1990s onward reported greater autonomy support for daughters than for sons (B = 0.01, 95% CI [0.00, 0.01], p < .01). In addition, across all studies, parents showed more autonomy support for girls than for boys at the youngest age level (0–2 years). Not yet known, however, is whether mothers and fathers of infants are equally likely to provide greater autonomy support for daughters than for sons.

In sum, the current study applied parallel home observations of mothers’ and fathers’ autonomy support for 14-month-old infants in structured dyadic play to (a) assess their similarity in form and mean levels and (b) examine individual differences in relation to parental and child factors (e.g., age, personality temperament, gender).

Procedure

The National Health Service (United Kingdom) Research Ethics Committee study approved the protocol. In the last trimester, expectant parents completed an online questionnaire and in-person interview. Parents completed online questionnaire and 1-hr home visits at 4 months (not described here) and 14 months. At 14 months each parent was observed interacting with the infant in structured play followed by a free-play task (with mother–father order counterbalanced across visits and infants given rest breaks if needed).

Parental autonomy support

Mothers and fathers were filmed with their child in separate 4-min dyadic sessions involving one of two Melissa & Doug eight-piece inset jigsaw puzzles (“Farm Friends”/“Vehicles” Large Peg Puzzle; see the online supplemental materials ), counterbalanced across parents. Researchers asked parents to work with their children to complete the task and to bring them back to the activity if they went off task. To minimize distraction, the researcher left the room during the interaction. The Autonomy Support Coding manual ( Whipple et al., 2011 ) was used to code each video. Parents were rated on a scale from 1 ( not autonomy supportive ) to 5 ( very autonomy supportive ). This indicated the degree to which they (a) provided appropriately tailored help (Concern for Competence); (b) used hints, instructions, and encouragement (Verbalisations); (c) kept their child on-task (Flexibility and Perspective Taking); and (d) involved the child as an active participant (Following Child’s Pace and Providing Opportunities for Choice). Supporting the validity of this rating procedure, Autonomy Support ratings have shown positive correlations with indicators of parenting quality, such as parental mind–mindedness, attachment representations, and sensitivity (e.g., Bernier et al., 2010 ; Bernier, Matte-Gagné, Bélanger, & Whipple, 2014 ) but have shown unique predictive associations with children’s later cognitive and social development (e.g., Bernier et al., 2012 ; Matte-Gagné, Bernier, & Lalonde, 2015 ). An experienced doctoral researcher trained three graduate researchers on a subsample of 20 videos. Note that the Autonomy Support coding scheme does not have an international reliability set, and the authors do not provide training. Instead, the coding team was trained against the lead coder’s ratings. The 20 training cases were selected by the lead coder to capture a wide range of performance. After detailed discussion and feedback on these cases, the four coders independently coded a reliability set of 15 mothers and 15 fathers. Interrater reliability was acceptable for each domain of autonomy support ratings (.70 < intraclass correlation < .81). Throughout the coding process, cases were regularly selected for discussion with the lead coder to prevent coder drift. Coders never coded mothers and fathers from the same family and were unfamiliar with the parents and children in each video.

Parental personality

Mothers and fathers completed the Ten Item Personality Inventory (TIPI; Gosling, Rentfrow, & Swann, 2003 ). Although the TIPI has typically exhibited poor internal consistency because there are only two 1- to 7-point items in each dimension, test–retest reliability over a 6-week period has been excellent ( Gosling et al., 2003 ). Moreover, supporting its validity, the TIPI subscales have shown strong correlations ( r s > .65) with subscales on lengthier personality questionnaires ( Gosling et al., 2003 ). Averaging across the two items for each dimension yielded five 1- to 7-point scores representing Extraversion, Agreeableness, Conscientiousness, Emotional Stability, and Openness to Experience.

Child characteristics

Mothers completed the 36-item Early Childhood Behavior Questionnaire (ECBQ; Putnam, Gartstein, & Rothbart, 2006 ) to provide 1- to 7-point ratings on three dimensions of temperament: Negative Affect, Surgency, and Effortful Control. Ratings from the 12 pertaining to Negative Affect (e.g., “How often did your child become sadly tearful?”), Surgency (e.g., “How often did your child seem full of energy, even in the evening?”), and Effortful Control (e.g., “How often did your child wait patiently?”) were averaged, excluding items rated as not applicable. Negative Affect scores showed low internal consistency (α = .26) and so were dropped from analyses. Scores for Surgency and Effortful Control showed moderate internal consistency (α = .66 and .63, respectively).

Analytic Strategy

Using a latent variable modeling framework in Mplus ( Muthén & Muthén, 2017 ), we first examined the latent factor structure of ratings for mothers’ and fathers’ autonomy support. Next, using nested model comparisons, we examined the measurement invariance of each latent factor across mothers and fathers before testing latent mean differences between mothers and fathers ( Brown, 2015 ). We then examined covariates of maternal and paternal autonomy support using multiple indicators, multiple causes models ( Brown, 2015 ; Kline, 2012 ). Because the ratings had nonnormal distributions, we used a robust maximum likelihood estimator for each of our models ( Kline, 2012 ) and three standard criteria: comparative fit index (CFI) of >.90, Tucker–Lewis index (TLI) of >.90, and a root-mean-square error of approximation (RMSEA) of <.08 ( Brown, 2015 ) to evaluate model fit.

Almost all mothers ( N = 194) and most fathers ( N = 188) had complete and valid observational data. Missing questionnaire data did not exceed 5% for mothers or 10% for fathers. The electronic format prevented nonresponse to individual items. To avoid loss of data, we used a full information approach so that all families who participated in the 14-month home visit ( N = 195) could be included in the analysis. Model parameters and standard errors were estimated in Mplus using all available data ( Enders, 2011 ).