Infant Emotion Regulation and Cortisol Response during the First Two Years of Life: Association with Maternal Parenting Profiles

Abstract

The current study investigated the prospective associations among emotion expression, behavioral regulation and cortisol responses in relation to different maternal parenting behaviors during the first two years of the infant’s life, among a sample of low-income families. Participants included 1141 mother-child pairs, assessed when the infants were 6-, 15-, and 24-month old. Maternal parenting behaviors were observed at the 6-month assessment, whereas infant emotion expression, orienting towards mothers, and cortisol responses were measured using a series of emotion-eliciting tasks at all time points. A Latent Profile Analysis revealed four maternal parenting profiles: Detached, Intrusive, Average and Engaged. Further, a multiple-group path model revealed distinct patterns of emotion development for infants within different maternal parenting groups. Among children with Engaged and Average mothers, orienting behaviors tended to predict less negative emotion and cortisol responses, which was associated with more future orienting behaviors. Conversely, among children with Intrusive and Detached mothers, orienting behaviors tended to predict more negative emotion and cortisol responses, which predicted less future orienting behaviors. Findings of this study enhance current understanding of how different profiles of maternal parenting behaviors impact infant emotional development in poverty, with significant implications for intervention programs targeting early mother-infant interactions.

Children’s ability to regulate emotion during infancy is associated with later socioemotional adjustment (Morris, Criss, Silk, & Houltberg, 2017). Emotion regulation (ER) refers to “the process of modulating the occurrence, duration, and intensity of internal states of feeling (both positive and negative) and emotion-related physiological processes” (Morris et al., 2017, p. 233). These processes can include three distinct but closely related aspects: emotion expression, behavioral regulation, and physiological regulation (Cole, Martin, & Dennis, 2004; Blair et al., 2011; Feldman, 2009). During infancy, these regulatory processes develop within the family context, in which mothers and other primary caregivers play a crucial role, because infants commonly rely on caregivers to provide support for regulation of emotion (Eisenberg, Cumberland, & Spinrad, 1998; Morris et al., 2017). Maternal parenting behavior is likely to influence the ways in which infants engage their mothers to facilitate behavioral regulation, which in turn will affect other aspects of ER (e.g., Ekas, Braungart-Rieker, Lickenbrock, Zentall, & Maxwell, 2011; Khoury et al., 2016). The current study investigated the longitudinal associations among emotion expression, behavioral regulation, and cortisol processes in the context of differential maternal parenting behaviors in the first two years of the infant’s life. Understanding the interrelations and co-development of the three aspects of ER in a sample of predominantly low-income families may be especially important, since a large body of literature has shown that young children growing up in poverty face increased risk for early emotional and behavior problems through dysregulated emotions (e.g., Blair et al., 2008; Raver, Blair, Garrett-Peters, & FLP Key Investigators, 2015).

ER behavior and Cortisol Response in Infancy

Feldman (2009) proposed a developmental hierarchical-integrative perspective to understand the development of early regulatory processes across the first several years of children’s life. This model suggests that the investigation of ER during the first two years of an infant’s life requires an understanding of the developmental associations among the physiological, emotional, and behavioral processes of ER (Feldman, 2009). More specifically, from a dynamic systems perspective (Lewis, 2005), ER is an integration of multiple components, and the process of ER development is dependent on synchronizing physiological, emotional, and behavioral aspects of ER as well as acquiring personalized patterns of responses through mutual influences among the three interrelated processes. With the maturation of each system as well as emerging patterns of interconnections among systems, these subcomponents of ER are organized into stable, individualized regulatory styles. Thus, an investigative approach centering on how individuals develop stabilized but different regulatory patterns is largely needed, particularly concerning the longitudinal feature of ER development. Finally, as early ER processes are highly plastic and malleable to changes in the environment, research on the development of ER calls for an understanding of contextual influences, particularly familial and parental behaviors (Feldman, 2009; Hostinar & Gunnar, 2013; Morris et al., 2017).

Significant development occurs during infancy in terms of ER skills. One of the important behavioral strategies of infants’ ER is orienting or displaying social signals towards the caregivers, usually their mothers, as they recognize that the caregivers may assist them in the regulation of affective states (Diener, Mangelsdorf, McHale, & Frosch, 2002). With the maturation of attention systems, infants before age one are able to use orienting behaviors to redirect attention and gain social support (Diener et al., 2002; Ekas, Braungart-Rieker, Lickenbrock, Zentall, & Maxwell, 2011). Successful use of such behaviors reduces negative emotion and facilities children’s transitioning from passive, caregiver-directed regulation to active self-regulation (Calkins & Hill, 2007).

Physiological regulation is another important aspect of infant’s ER. Studies on physiological regulation have frequently focused on the glucocorticoid hormone cortisol, usually obtained from saliva samples. Cortisol is the end product of activities in the hypothalamic-pituitary-adrenal (HPA) axis, which is highly reactive to external stressors during children’s development (Hostinar & Gunnar, 2013). The activity levels of the HPA axis can indicate an infant’s levels of reactivity and regulation (Ahnert, Gunnar, Lamb, & Barthel, 2004; Khoury et al., 2016). The changes in cortisol levels in times of distress are indicative of HPA regulation of stress. As such, a typical approach to examine HPA regulation is through salivary cortisol responses to stress, which usually show an increase after challenge followed by a drop in cortisol levels, normally occurring approximately 30-40 minutes later (Blair et al., 2011). To capture cortisol reactivity and recovery, saliva samples are usually obtained before the challenge task and several times after the stressor, within a 40-minute timeframe. Changes from pre- to post-challenge are commonly used to define HPA reactivity. A higher level of increase in cortisol levels after stress indicates higher HPA reactivity and less successful regulation (Laurent, Harold, Leve, Shelton, & Van Goozen, 2016). On the other hand, a larger decrease in cortisol levels post stress indicates greater success in HPA regulation (Albers et al., 2008; Schoorl et al., 2016).

The HPA axis is highly responsive to stressors at birth; newborns show significant elevations in saliva cortisol even to minor disruptions (Gunnar et al., 2007). By six months of age, the HPA system of the human child is relatively mature (Loman & Gunnar, 2010). Infants’ negative emotions, such as fussing and crying, are closely related to their cortisol reactivity in the first year (Ahnert et al., 2004; Lewis & Ramsay, 2005; Ursache et al., 2013). By the end of the first year of life, the human HPA system becomes less responsive to stressors, and researchers observed a separation between behavioral distress and cortisol responses among 15-month-old infants compared to infants less than one year old (Gunnar et al., 1996). From the second year and on, small increases in cortisol levels are observed in response to experimentally induced stressors (Blair et al., 2005), and individual variations become more significant (Blair et al., 2007; Thompson et al., 2015).

The HPA axis is one of the underlying physiological systems of stress regulation. Compared to behavioral responses, HPA axis usually reacts to stressors at a slower pace but over time exhibits a more stable pattern signaling one’s regulatory capacities (Thompson, Morgan, & Jurado, 2015). As the malleability of HPA regulation is very high during infancy and gradually reduces as infants age, Thompson et al. (2015) suggested that cortisol responses can be used as an indicator of the formation of a stable, individualized regulation pattern. That is, once infants establish emotional and behavioral regulatory patterns, their HPA reactions are also stabilized, displaying consistent patterns of responding to external and internal stress. In this way, negative emotion and cortisol responses would mutually influence each other during infancy, as the former indicates observable, affective process of emotional responses, and the latter presents a rather stable pattern of physiological reactivity and regulation to external stimuli.

Maternal Influences on Early Emotional Regulation

Attachment theory (Bowlby, 2012) delineates how maternal assistance influences infants’ regulation, suggesting that several types of early maternal parenting behaviors are crucial for child emotional outcomes. The first one is maternal sensitivity, the ability to perceive and interpret infants’ emotional signals accurately and to respond to them promptly and adequately (Ainsworth, Blehar, Waters, & Wall, 1978). As attachment processes are activated in emotion-eliciting contexts and serve specific ER functions, sensitive parenting contributes to the acquisition of ER skills in infancy and toddlerhood. Maternal sensitivity to their infant’s distress was associated with fewer behavioral problems and higher social competence in toddlerhood, especially for children high in negative emotionality (Leekers et al., 2009).

In contrast, insensitive parenting, intrusiveness and detachment, is associated with negative child emotional outcomes. Intrusive parenting includes negativity in affective expressions and interference of child activities. Detached mothers, by contrast, are usually unresponsive, affectively flat, and barely supportive towards their infants’ activities (Tronick & Reck, 2009). Intrusive and detached parenting is associated with increased negative emotion and decreased positive affect over the first two years of infants’ life (Field, 2010; Kochanska, 2001). Maternal intrusiveness also predicted increased child cortisol reactivity from age one to three (Laurent et al., 2016). Detached parenting during infancy predicted higher rates of toddler behavioral problems (Jones Harden, Denmark, Holmes, & Duchene, 2014).

Maternal parenting is also related to infant ER behaviors. Typically, infants signal and elicit support from the primary caregiver at times of distress through behaviors such as looking, crying, and clinging to others, which serve a survival purpose and fulfill their attachment needs (Bowlby, 2012; Cassidy, 1994). With sensitive mothers, infants can safely express both positive and negative emotions. Conversely, infants of insensitive mothers either use heightened distress to attract their mothers’ attention or display a flat affect to maintain the needed proximity without threatening the mother-child relationship. Supporting this, securely attached infants use social orientation strategies, whereas insecure/avoidant children use self-soothing and solitary exploration with toys (Diener et al., 2002; Ekas et al., 2011). The ER strategies of infants reflect their expectations about the availability of the caregiver in assistance of ER.

Although most previous studies classified maternal parenting styles by observation, they did not consider a person-centered approach that would allow for the investigation of different constellations of maternal parenting behaviors during infancy. A person-centered approach can capture different patterns of associations among multiple parenting constructs to identify distinct group-level characters; further, it would help the exanimation of child outcomes predicted by multiple parenting behaviors simultaneously, while considering possible interactions among parenting variables (Lanza & Cooper, 2016). A person-centered approach provides a valuable framework for examining differential developmental trends that occur within distinct parenting profiles. Several studies have applied a person-centered approach to the study of maternal characteristics associated with emotional development among toddlers and preschoolers (e.g., Brophy-Herb et al., 2013; Hooper et al., 2015; Nelson et al., 2012), suggesting that the influences that mothers have on their children’s emotional development are built on the combination of several maternal parenting characteristics.

Associations among Aspects of Infant ER across Maternal Parenting Profiles

Several recent studies indicate that the associations among infant behavioral regulation, negative emotion and cortisol responses need to be understood while considering maternal influences. That is, under varying levels of maternal parenting quality, infant’s ER strategies are associated with distinct regulatory outcomes. For example, Ekas et al. (2011) found that orienting to an unresponsive caregiver increases subsequent negative emotion expressions in 20-month-old infants, whereas orienting to a responsive caregiver reduces negative emotion. Further, Khoury and colleagues (2016) studied a community sample of mothers who self-reported their depressive symptoms, as depressed mothers are likely to exhibit intrusive or detached parenting behaviors. Among mothers with elevated depressive symptoms, infant’s using solitary ER strategies (such as distraction or orienting to an object) within the mother’s presence are associated with higher cortisol reactivity, showing a maladaptive regulatory pattern. Similarly, when separated from mothers, higher cortisol reactivity was associated with more concurrent self-soothing behaviors among securely attached infants, but not insecure infants. The authors suggested that physiological stress responses among securely attached infants can help signal distress and elicit external assistance from their mothers (Beijers et al., 2017). Together, these aforementioned studies seem to suggest that as infants start to establish expectations about maternal sensitivity and responsiveness through early interactions with their mothers, utilizing orienting strategies can lead to different emotional and physiological outcomes between infants with sensitive mothers and those with intrusive or detached mothers. However, research evidence remains insufficient concerning how subtypes of insensitive mothers (e.g., intrusive versus detached mothers) differ, and how this may affect their infants’ emotional development.

SES, Race and Gender Influences

Research on maternal parenting and infant ER in low-income samples has been scarce, as most studies focus on the effect of early adversity on later development, such as middle childhood, adolescence, and adulthood. Studies show that young children in low-income families exhibit different patterns of cortisol responses compared to those of higher socio-economic levels, such as a higher level of resting cortisol in infants and toddlers (Blair et al., 2011) and a flatting cortisol trajectory in response to stress among four-year-olds, indicating a less optimal ability to regulate (Blair et al., 2013). Preliminary evidence also reveals infants using their mothers as sources of ER among low-income families. For example, among low-income families, one-year-olds show a higher ability to regulate emotions when mothers were emotionally available and less hostile (Little & Carter, 2005). Infants relying on their mothers for emotion challenges displayed higher cognitive and language skills at age 2 compared to those who did not (Robinson & Acevedo, 2001). Maternal cognitive stimulation was associated with better toddler language skills through infant emotional and eye-gaze communication (Cates et al., 2012). Finally, a few studies suggest that maternal sensitivity may be compromised by poverty, and mothers can still vary in their parenting behaviors in low-income, minority samples (e.g., Bakermans-Kranenburg, van IJzendoorn, & Kroonenberg, 2004; Ispa et al., 2004; Mesman, van IJzendoorn, & Bakermans-Kranenburg, 2012). However, there is a lack of studies differentiating the distinct effects of maternal sensitivity, intrusiveness, and detachment in infancy, and associated ER performance, especially in a low-income population.

The current study also oversampled minority families. Previous research revealed that African American mothers were less likely to encourage negative expressions in their children, and their encouragement was associated with less emotion competence in children, compared to Causation mothers. This observation is potentially due to the hostile climate of prejudice that this population has experienced (Hooper, Wu, Ku, Gerhardt, & Feng, 2018; Nelson et al., 2013). Consequently, their children are likely to show less negative emotion, as well as lower cortisol reactivity, than Caucasian infants (Blair et al., 2008).

The effect of child gender was also investigated, as it is commonly related to maternal parenting and child emotional adjustment (Chaplin, 2015). Mothers tend to be more sensitive towards girls’ negative emotion, based on their gender socialization beliefs in which girls are allowed to show more negative emotion and get parental comfort, compared to boys (Chaplin, 2015). As such, infant girls show earlier abilities to regulate emotion than boys (Thompson et al., 2015), and their use of effective regulatory strategy (i.e., distraction) is associated with less negative expression later (Crockenberg, Leerkes, & Jó, 2008). However, research attempts on early differences in ER considering infant gender and race are still limited.

The Current Study

Given the paucity of research with low-income families during the first two years of the infant’s life, the current study expanded on previous studies in attempting to unravel the relations among early maternal parenting, child emotion expression, behavioral regulation, and cortisol responses in a low-income sample. As we were interested in how infants develop individualized regulatory patterns associated with early maternal parenting, we used a person-centered approach and identified different parenting profiles observed from a mother-child interaction task. We further investigated the relations among child emotional, behavioral and cortisol responses to distressing situations within different maternal parenting profiles. We were interested in young children aged from 6 to 24 months, as they are highly susceptible to maternal influence and largely dependent on caregivers to regulate emotions (Calkins & Hills, 2007).

It was first hypothesized that different parenting profiles would exist in this low-income sample, and we expected to find profiles of sensitive, intrusive, and detached mothers, respectively, similar to observations of middle- and upper-classes of mothers (e.g., Bowlby, 2012; Field, 2010; Tronick & Reck, 2009). Second, maternal parenting profiles would moderate the prospective associations among child orienting behaviors, negative emotion, and cortisol responses. As suggested by the developmental hierarchical-integrative perspective (Feldman, 2009; Thompson et al., 2015), greater cortisol reactivity and poorer cortisol recovery in response to stress would predict higher child negative emotion, and higher child negative emotion would predict greater child cortisol reactivity and poorer cortisol recovery, likely for all children. However, for children with sensitive mothers, elevated negative emotion and cortisol reactivity and a greater cortisol recovery would be associated with more child orienting behaviors, and child orienting behaviors would be associated with less negative emotion, lower cortisol reactivity and a greater cortisol recovery. For children having intrusive and disengaged mothers, elevated negative emotion, greater cortisol reactivity, and less cortisol recovery would be associated with fewer child orienting behaviors, and more child orienting behaviors would be associated with higher levels of negative emotion, greater cortisol reactivity, and less cortisol recovery among these families.

Method

Participants

Data of the current study were drawn from the Family Life Project (FLP; Vernon-Feagans & Cox, 2013). The FLP was designed to study the development of young children and their families in two of the four major geographical areas of the United States with high poverty rates. The FLP included a sample of 1,292 children and oversampled low-income and African American families. Recruitment spanned from September 15, 2003 to September 14, 2004, when these families completed a home visit at 2 months of child age (see Vernon-Feagans & Cox, 2013 for details of study design and recruitment). The current study included 1141 families that have completed the mother-child interaction task at the age 6 month visit (T1), and these families were assessed two more times at the child aged 15 months (T2; retention rate 85.9%) and 24 months (T3; retention rate 77.1%). The current sample did not differ from the original sample in terms of child gender, race, income level, maternal age, education, employment status, and marital status.

The sample included 576 boys (50.5%) and 565 girls (49.5%). Approximately forty percent of the sample were African American (40.3%), 59.3% were Caucasian, and 0.4% were other races. Almost half of the mothers were married (48.8%), 48.0% single, and 3.2% divorced/separated/widowed. Less than half of the mothers (40.2%) reported being employed, while the rest (59.8%) were unemployed at the time of study enrollment. As to the highest level of maternal education, 19.5% did not complete high school, 4.8% received a GED, 34.5% graduated from high school, 26.7% completed some college or had an associate’s degree, 9.6% had a college degree, and 4.8% had post-undergraduate training.

Procedure

Families participating in the study were visited at their homes for data collection at T1, T2, and T3 by trained research assistants. The child’s primary caregiver, in 99% of cases the mother, was administered questionnaires on demographic characteristics. Mothers engaged in a free-play interaction task for 10 minutes with their children at T1, which was recorded using a digital video (Cox, Paley, Burchinal, & Payne, 1999; NICHD ECCRN, 1999), and later coded for their parenting behaviors. See other studies for recruitment strategies and descriptions of participating families and their communities (Burchinal, Vernon-Feagans, Cox, and the FLP Investigators, 2008; Vernon-Feagans, Cox, and the FLP Investigators, 2013).

Children were also administrated several tasks eliciting negative emotion at each time point, with their mothers’ present. All these observation tasks were videotaped and later coded for children’s negative emotion and ER behaviors. At 6 months, three tasks were administrated: the arm restraint task (Stifter & Fox, 1990), the barrier task (Goldsmith & Rothbart, 1996), and the mask task (Goldsmith & Rothbart, 1996). In the arm restraint task, a research assistant stood behind the infant and gently held the infants’ arms for 2 minutes, then released his/her arms and stayed hiding behind the infant for 1 minute where the infant employed regulation behaviors prior to maternal comfort. In the barrier task, a research assistant gave the child an attractive toy to play for 30 seconds, after which the experimenter gently removed the toy and placed it behind a Plexiglas barrier in front of the infant for 30 seconds. The procedure was repeated twice more for 3 complete trials. In the mask task, the experimenter sat to the side of the child and put on four consecutive scary masks, each for 10 seconds. Each time, the experimenter would speak the child’s name while moving her head slowly from side to side and then lean toward the child.

At 15 and 24 months, two tasks were administrated to children: the mask task and the toy removal task (Goldsmith & Rothbart, 1996). In the toy removal task, for the first 2 minutes, infants and their mothers were presented with an attractive toy and asked to play together to get the child engaged. In the 2 minutes following, the toy was taken away from the child and placed out of their reach (T2) or in a clear plastic jar (T3). Finally, in the last minute, the toy was returned to the child. The infant’s negative emotion and behavioral regulation when the toy was taken away were used in the current study.

At home visits for data collection at T1, T2, and T3, saliva samples were collected from children prior to the emotion-inducing tasks, and 20-minute and 40-minute after these tasks. Home visits were scheduled when families were available, and time of the day of saliva collection varied according to the length of each visit and the family/infant availability. Mean time of day of saliva sample collection was 13:04 hours (SD = 2.88) at T1, 13:45 hours (SD = 2.94) at T2, and 13:33 hours (SD = 3.20) at T3. The number of families that had available cortisol data were 1082, 945, and 889 at T1, T2 and T3, respectively. Reasons for missing cortisol data were primarily due to insufficient volumes of saliva for assays or participant refusal. Procedures for handling missing data are described in the “Analytic Strategies” section. The University’s Institutional Review Board approved all study procedures.

Measures

Maternal parenting was assessed using a mother-child free play task at T1. Seven maternal behaviors were coded during this interaction: sensitivity (how the mother observes and responds to the child’s social gestures, expressions, and signals as well as responds to cries, frets, or other expressions of negative affect); intrusiveness (adult-centered interaction, imposing their agenda on the child); detachment (emotionally uninvolved and unaware of the child’s needs for appropriate interaction to facilitate involvement with objects or people); positive regard (positive feelings toward the child); negative regard (negative feelings toward the child); animation (how animated the mother is); and simulation of development (trying to foster the child’s development). Ratings for the codes were made on a 1–5 scale, with one being not at all characteristic and five being highly characteristic. Approximately 50% of the task was double-coded, with reliability kappa values ranging from 0.83 to 0.88 for all codes.

Child negative emotion was assessed using a series of emotion-eliciting tasks at each time point (the arm restraint task, the barrier task, and the mask task at T1; the mask task and the toy removal task at T2 and T3). A coding system was adapted (Stifter & Braungart, 1995) to code child negative emotion, including behaviors such as crying, screaming, whining, and frowning. Child negative emotion was coded second-by-second, on a four-point scale: 0 = no negative, 1 = low negative, 2 = moderate negative and 3 = high negative. A weighted sum score was computed by summing up the proportion of time in each level of negative emotion multiplied by weight (i.e., no negative *0, low negative *1, moderate negative *2 and high negative *3). The scores across different tasks at each wave were standardized and summed up to indicate the child overall negative emotion. Approximately 15% of each task was double-coded, and the reliability of the codes varied between 0.85 and 0.95.

Child orienting to mothers was assessed from the same emotion-eliciting tasks at each time point. A coding system (Stifter & Braungart, 1995) was used to code the duration of time that the child was looking at the mother and communicating nonverbally with body languages. The behavior was coded second-by-second for each task, and a final score is generated to represent the percentage of time that the orienting behavior was present (out of the total coded duration of the task). The scores across different tasks at each wave were standardized and summed up to indicate the child overall tendency to use orienting behaviors. Approximately 15% of each task was double-coded, and the reliability of each code varied between 0.82 and 0.95.

Child salivary cortisol.

Saliva samples were assessed for cortisol using a highly sensitive enzyme immunoassay US FDA 510k cleared for use. The test used 25 μl of saliva, with a range of sensitivity from .007 to 1.8 μg/dl and average intra-and inter-assay coefficients of variation of less than 10%. All samples were assessed twice. The criterion for repeat testing was variation between duplicates greater than 20%, and the average of the duplicates was used. There were no relations between child temperature, time since eating, time since sleeping, and use of medications (e.g., acetaminophen) and child cortisol levels. At each time point, child cortisol was assessed three times: prior to the emotion-eliciting tasks (baseline), 20 minutes after the tasks (peak) and 40 minutes after the tasks (post-stress rest period). Child cortisol reactivity was calculated by subtracting baseline cortisol from the peak cortisol; cortisol recovery was computed by subtracting the post-stress rest period cortisol from the peak level, indicating successful regulation (Albers et al., 2008; Beijers et al., 2017).

Covariates included child sex, race and family income-to-needs ratio. Child sex and race was reported by mother when the child was 2-month-old, and family income-to-needs ratio was assessed at T1, by dividing total family income by the poverty threshold for the appropriate family size, with an average of 1.82 (SD = 1.70, range 0.00-16.49).

Analytic Strategies

Analysis was conducted using the Mplus 8 program (Muthén & Muthén, 2017). The root mean squared error of approximation (RMSEA) and comparative fit index (CFI) were used to evaluate the model fit, with a RMSEA of 0-.05 and CFI of .95-1.00 indicating good fit and a RMSEA of .05-.08 and CFI of .90-.95 showing acceptable fit (Hu & Bentler, 1995).

To test the first hypothesis of different maternal parenting profiles, a Latent Profile Analysis (LPA) was conducted based on the seven indicators coded from the mother-child interaction task at T1. A series of LPA models with different numbers of classes (2-6 classes) were estimated to determine the model that best fitted the data. Several criteria were used to determine the best fitting model: a) Bayesian Information Criteria (BIC); b) the p-value of the Vuong-Lo-Mendell-Rubin (VLMR) test, c) the p-value for Bootstrapped Likelihood Ratio Test (BLRT), d) a reasonable number of mothers assigned into each class, and e) interpretability of the class memberships (Jung & Wickrama, 2008). Preliminary results concerning the mean differences in demographic variables across parenting profiles were tested by the automatic three-step approach estimating predictors to latent classes in LPA (i.e., the R3STEP option in Mplus). Regarding the mean differences of child variables across parenting profiles, we used the automatic three-step approach estimating distal outcomes in LPA (i.e., the DCON option in Mplus) as recommended (Asparouhouv & Muthén, 2013). To avoid possible inflation of Type I errors caused by multiple comparisons, we applied a correction to the p-value and elected to only report those findings with a significance level of 0.01 or stronger. The assigned groups were then used as observed variables in conducting the multiple-group analysis.

To test the second hypothesis about the associations among orienting, negative emotion and cortisol responses across distinct parenting profiles, a multiple-group path model was estimated with auto-regressive, cross-lagged paths among child cortisol, negative emotion and orienting to mothers, to investigate the differences in the cross-lagged paths among the parenting profiles. To test this, two models were estimated, one without constraints and another with constraints on the cross-lagged pathways among variables being the same across the parenting profile groups. The chi-square difference test was used to compare the model fits of the two nested models. If the unconstrained model yielded a better model fit than the constrained model, the unconstrained model would be retained and the target pathways would be left unconstrained, as this indicates the parameter estimates of the target pathways differed significantly across groups. Otherwise, if the unconstrained model did not show a better model fit than the constrained model, the constrained model would be kept with the target pathways being equal across groups. Between-group comparisons on the pathways which were significantly different across groups were conducted, by testing the pairwise differences on the paths across groups. Child sex, race and family income-to-needs ratio were included as covariates in the path model.

Little’s MCAR test (Little, 1988) revealed that data were not missing completely at random, χ²(2810) = 3025.84, p < 0.01. However, it is reasonable to expect that the data were missing at random (MAR), as missingness of data was associated with other study variables. Missingness of cortisol variables was associated with child negative emotion (rs = 0.06-0.18, ps < 0.01) and orienting to mothers (rs = 0.06-0.08, ps <0.05). Missingness of negative emotion was associated with child negative emotion at different time points (rs = 0.07-0.09, ps < 0.05) and orienting to mothers at T1 (rs = 0.07-0.08, ps <0.05). Missingness of child orienting to mothers at T1 was associated with child negative emotion at T2 (r = 0.09, p < 0.01), income-to-needs ratio (r = −0.07, p = 0.03) and maternal intrusiveness (r = −0.07, p = 0.03), and missingness of child orienting at T3 was associated with child orienting at T1 (r = 0.08, p = 0.01). Full information likelihood (FIML) estimation was thus used as recommended to generate unbiased estimates under the condition of MAR (Enders & Bandalos, 2001). Children’s observed negative emotion and orienting mother scores were positively skewed (Table 1). Maximum likelihood estimation with robust (Huber-White) standard errors was used to estimate the model, as recommended for non-normal distributions (Li, 2015).

Table 1.

Descriptive Statistics and Bivariate Correlations of Study Variables.

		1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21	22
1.	Child Sex
2.	Ethnicity	.00
3.	Income	−.04	.38***
Maternal parenting T1
4.	Sensitivity	.06*	.37***	.33***
5.	Intrusiveness	−.05	−.28***	−.25***	−.54***
6.	Detachment	−.03	−.29***	−.26***	−.69***	.04
7.	Stimulation	.04	.24**	.27***	.48***	−.08**	−.58***
8.	Pos regard	.01	.34***	.29***	.58***	−.13***	−.67***	.55***
9.	Neg regard	−.01	−.31***	−.23***	−.31***	.47***	.07*	−.08**	−.15***
10.	Animation	−.01	.28***	.27***	.51***	−.02	−.68***	.59***	.74***	−.01
Child cortisol
11.	Reactivity T1	.05	.03	.02	.05	−.01	−.06	.04	.07*	.03	.04
12.	Reactivity T2	.02	.08*	.02	.00	−.03	.03	−.02	−.04	−.05	−.03	−.02
13.	Reactivity T3	.00	−.04	−.04	.02	.01	−.01	.01	.01	.03	.04	.02	.08
14.	Recovery T1	−.01	.04	.03	.04	−.01	−.04	.04	.03	.01	.00	.70**	.04	.03
15.	Recovery T2	.00	.02	−.02	.01	−.06	.02	−.02	.01	−.03	.02	−.05	.67**	.02	−.03
16.	Recovery T3	.04	−.02	−.04	.00	.03	−.01	.01	−.02	.05	.03	.00	.06	.76**	.03	.03
Child regulation
17.	Neg Emo T1	.09**	.02	.04	.05	−.05	−.03	−.03	.07*	−.03	.05	.08*	.02	.03	.01	.02	.06
18.	Neg Emo T2	.00	−.16***	−.05	−.06	.08**	−.01	−.05	−.03	.07*	.01	.08*	.06	.12**	.07	.03	.10**	.16***
19.	Neg Emo T3	−.09**	−.06	−.03	−.03	.03	.02	−.02	−.04	.01	.00	.04	.03	.12**	.04	.03	.04	.04	.29***
20.	Orienting T1	.13***	−.09**	−.02	.00	.04	−.06*	.00	−.01	.04	.04	.10**	.03	.00	.01	.00	.02	.11***	.04	.00
21.	Orienting T2	.05	.22***	.09**	.13***	−.08*	−.14**	.12***	.07*	−.09**	.06	.02	.06	−.01	.01	−.01	.01	.01	.15***	.01	.02
22.	Orienting T3	.03	.03	.07*	.08*	−.02	−.03	.02	.01	.04	.04	.00	.07	.02	−.01	.03	.07	.07*	.08*	.18***	02	.11**
N		1141	1141	1141	1141	1141	1141	1141	1141	1141	1141	970	851	832	902	786	810	1036	972	866	1042	980	880
Mean		--	--	1.82	2.68	2.87	2.81	2.65	2.92	1.95	3.04	0.01	0.03	0.00	−0.01	0.01	−0.00	0.01	0.00	0.01	0.00	0.00	0.00
SD		--	--	1.70	0.83	0.82	1.00	0.99	0.98	0.96	0.96	0.25	0.33	0.21	0.27	0.32	0.25	1.99	1.51	1.47	1.74	1.44	1.45
Min		1.00	0.00	0.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00	−2.07	−2.48	−1.38	−1.87	−2.11	−1.35	−2.03	−2.10	−1.71	−2.32	−1.86	−1.36
Max		2.00	1.00	16.49	5.00	5.00	5.00	5.00	5.00	5.00	5.00	1.39	2.43	1.77	1.86	1.89	1.36	9.27	5.23	5.68	12.12	8.73	11.57
Skewness		0.02	−0.30	2.51	−0.20	0.15	0.30	0.42	−0.13	0.92	−0.02	−1.39	−1.25	0.58	0.49	−0.31	−0.75	1.25	0.64	1.00	1.78	1.37	2.23

Note.

^*p < .05,

^**p < .01,

^***p < .001.

Child Sex 1= Boy, 2 = Girl. Ethnicity 0 = Black, 1 = Other races. Pos = positive, Neg = negative, Emo = emotion.

Results

Descriptive statistics and bivariate correlations of study variables are shown in Table 1. Maternal sensitivity at T1 was positively associated with child orienting mothers at T2 and T3. Maternal intrusiveness and negative regard were related to elevated child negative emotion and less orienting at T2. Maternal detachment was related to less child orienting at T1 and T2. Maternal stimulation was positively associated with child orienting at T2. Maternal positive regard was associated with greater concurrent child cortisol reactivity and negative emotion. Child negative emotion at T1 was related to greater cortisol reactivity at T2. Child orienting at T1 was associated with greater cortisol reactivity at T1, whereas child orienting at T1, T2 and T3 were associated with elevated concurrent negative emotion.

Maternal Parenting Profiles

For the first hypothesis, a series of Latent Profile Analyses were used to identify subgroups of mothers, based on the similarity of observed maternal parenting variables (sensitivity, intrusiveness, detachment, positive regard, negative regard, and stimulation of development) at T1 (Table 2). The VLMR test showed that the 4-class solution was statistically better than the 3-class solution, and the 6-class solution was better than the 5-class solution. Though the 6-class solution has the smallest BIC value, there was one group comprised by less than 3% of the sample, which is usually not recommended (Jung & Wickrama, 2008). Based on the theoretical hypothesis and the fit indices, the 4-class solution was selected. As shown in Figure 2, significant differences among means of parenting variables were found. Based on characteristics of each parenting profile, these groups were labeled as Detached (n = 236, 20.7%; high on detachment, average on intrusiveness and negative regard, and low on other variables); Intrusive (n = 171, 15.0%; high on intrusiveness and negative regard, average in detachment, stimulation, positive regard and animation, and low on sensitivity); Average (n = 428, 37.5%; average on all variables); and Engaged (n = 306, 26.8%; high on sensitivity, stimulation, positive regard and animation, low on intuitiveness, detachment and negative regard). The categorical latent variable, the group membership, was saved to be used in the following analysis.

Table 2.

Fit statistics for LPA models with 1- to 6-class models

LPA models	Class sizes	Entropy	BIC	Adjusted BIC	VLMR p value	BLRT p value
1-class	1141	--	21654.030	21609.561	--	--
2-class	372,769	0.852	19606.376	19536.497	0.00	0.00
3-class	226,538,377	0.837	18845.036	18749.747	0.00	0.00
4-class	236,428,306,171	0.88	18338.717	17979.322	0.01	0.00
5-class	101,177,569,151,142	0.985	16366.946	16220.836	0.50	0.00
6-class	71,569,169,151,151,30	0.99	15621.320	15840.654	0.00	0.00

Note. BIC = Bayesian Information Criteria; VLMR = Vuong-Lo-Mendell-Rubin; BLRT = Bootstrapped Likelihood Ratio Test; LPA = Latent Profile Analysis.

Figure 2.

Means of the parenting characteristics of each identified profile.

The three-step approach to LPA (Asparouhouv & Muthén, 2013) was used to examine the differences in other variables of interest among groups. Significant between-group differences were found for child ethnicity and income-to-needs ratio. The Engaged group showed a lower percentage of African American children, compared to the Average group (B = −1.00, SE = 0.22, t = −4.63, p < .001), the Detached group (B = −2.14, SE = 0.22, t = −9.68, p < .001), and the Intrusive group (B = −1.92, SE = 0.24, t = −8.06, p < .001); and a higher income-to-needs ratio, compared to the Average group (B = 0.31, SE = 0.07, t = 4.40, p < .001), the Detached group (B = 0.82, SE = 0.11, t = 7.31, p < .001), and the Intrusive group (B = 0.56, SE = 0.10, t = 5.79, p < .001). The Average group showed a lower percentage of African American children compared to the Detached group (B = −1.15, SE = 0.18, t = −6.34, p < .001) and the Intrusive group (B = −0.92, SE = 0.20, t = −4.61, p < .001), and a higher income-to-needs ratio compared to the Detached group (B = 0.51, SE = 0.11, t = 4.72, p < .001) and the Intrusive group (B = 0.25, SE = 0.09, t = 2.64, p = .008). There were no differences between the Detached and Intrusive groups regarding child ethnicity and income-to-needs ratio. We further compared the between-group differences of child regulation variables. The Average group showed higher cortisol reactivity than the Engaged group at T2 (χ²(1) = 9.42, p = .002). The Engaged group showed higher orienting behaviors at T2 than the Detached (χ²(1) = 13.58, p < .001), Intrusive (χ²(1) = 15.45, p < .001), and Average (χ²(1) = 7.02, p = .008) groups. The Intrusive group showed poorer cortisol recovery at T3 compared to the Average group (χ²(1) = 9.32, p = .002) and the Detached group (χ²(1) = 7.50, p = .006).

Multiple-Group Path Analysis

An autoregressive and cross-lagged path model was estimated to test the prospective associations among cortisol reactivity and recovery, negative emotion, and orienting to mothers, across three time points (Figure 1). First, all the paths were freely estimated across groups, and then individual paths were set to be equal across groups, respectively, to test if each path differs significantly among groups (Table 3). If the two models showed no statistically significant difference in model fit, the latter was adopted for parsimony. If the model fit became worse, the path was then set to be different among groups. The final model fits the data well, χ²(266) = 319.16, p = .01; RMSEA = .027 (CI_.90 = .013, .037); CFI = .958.

Figure 1.

The conceptual model.

Note. T1 = 6 month visit; T2 = 15 month visit; T3 = 24 month visit.

Table 3.

Coefficient Estimate for the Multiple-Group Model.

	Detached		Intrusive		Average		Engaged		Group
	β	t	β	t	β	t	β	t	Δχ2(3)
Cortisol reactivity T2	R2 = .01		R2 = .03		R2 = .01		R2 = .01
Cortisol reactivity T1	−0.02	−0.56	−0.01	−0.55	−0.02	−0.57	−0.02	−0.55	ns
Negative emotion T1	0.01	0.25	0.01	0.24	0.01	0.24	0.01	0.25	ns
Orienting T1	0.03	0.68	0.13	2.56**	−0.06	−1.32	0.05	1.14	12.05**
Child sex	0.01	0.27	0.01	0.27	0.01	0.27	0.01	0.27	ns
Ethnicity	0.09	2.92**	0.09	2.76**	0.10	3.47***	0.09	2.78**	ns
Income-to-needs ratio	−0.01	−0.75	−0.01	−0.74	−0.02	−0.75	−0.03	−0.74	ns
Cortisol reactivity T3	R2 = .02*		R2 = .01		R2 = .02		R2 = .03*
Cortisol reactivity T2	0.06	1.73	0.04	1.75	0.04	1.74	0.06	1.67	ns
Negative emotion T2	0.13	3.18***	0.09	2.83**	0.10	2.84**	0.15	3.30***	ns
Orienting T2	−0.03	−0.53	−0.02	−0.54	−0.02	−0.54	−0.03	−0.55	ns
Child sex	−0.01	−0.36	−0.01	−0.36	−0.01	−0.36	−0.02	−0.36	ns
Ethnicity	−0.02	−0.46	−0.02	−0.46	−0.02	−0.46	−0.02	−0.45	ns
Income-to-needs ratio	−0.04	−1.23	−0.03	−1.22	−0.04	−1.21	−0.07	−1.35	ns
Cortisol recovery T2	R2 = .01		R2 = .03		R2 = .01		R2 = .01
Cortisol recovery T1	−0.09	−1.57	−0.06	−1.62	−0.08	−1.40	−0.07	−1.48	ns
Negative emotion T1	0.03	0.73	0.02	0.69	0.03	0.71	0.02	0.71	ns
Orienting T1	−0.01	−0.41	−0.01	−0.41	−0.01	−0.40	−0.01	−0.41	ns
Child sex	0.02	0.50	0.02	0.50	0.02	0.48	0.02	0.50	ns
Ethnicity	0.04	1.20	0.03	1.14	0.04	1.22	0.02	1.17	ns
Income-to-needs ratio	−0.01	−0.20	−0.16	−1.65	−0.01	−0.32	−0.02	−0.61	9.11*
Cortisol recovery T3	R2 = .02		R2 = .00		R2 = .02		R2 = .02
Cortisol recovery T2	0.06	1.60	0.03	1.48	0.06	1.56	0.08	1.36	ns
Negative emotion T2	0.12	3.02**	0.05	2.58**	0.10	2.77**	0.11	2.68**	ns
Orienting T2	−0.05	−1.00	−0.02	−1.00	−0.04	−1.03	−0.05	−1.06	ns
Child sex	0.05	1.20	0.02	1.15	0.04	1.25	0.04	1.23	ns
Ethnicity	0.03	0.56	0.01	0.55	0.03	0.55	0.02	0.57	ns
Income-to-needs ratio	−0.04	−1.24	−0.02	−1.21	−0.05	−1.22	−0.06	−1.33	ns
Negative emotion T2	R2 = .06***		R2 = .12**		R2 = .08***		R2 = .06**
Negative emotion T1	0.17	4.98***	0.17	4.64***	0.19	4.84**	0.18	5.00***	ns
Cortisol reactivity T1	0.03	0.66	0.03	0.67	0.03	0.66	0.04	0.66	ns
Cortisol recovery T1	0.07	1.28	0.06	1.25	0.07	1.31	0.07	1.43	ns
Orienting T1	−0.01	−0.22	−0.01	−0.22	−0.01	−0.22	−0.01	−0.22	ns
Child sex	0.02	0.30	−0.25	−3.35***	0.05	1.02	−0.05	−0.92	15.40***
Ethnicity	−0.15	−4.50***	−0.15	−4.56***	−0.17	−4.71***	−0.14	−4.37***	Ns
Income-to-needs ratio	0.01	0.28	0.01	0.28	0.01	0.28	0.01	0.28	Ns
Negative emotion T3	R2 = .09***		R2 = .13***		R2 = .10***		R2 = .11***
Negative emotion T2	0.30	7.42***	0.32	7.11***	0.30	7.99***	0.30	8.39***	Ns
Cortisol reactivity T2	0.02	0.43	0.02	0.42	0.02	0.44	0.02	0.42	ns
Cortisol recovery T2	0.01	0.25	0.01	0.25	0.01	0.25	0.01	0.25	ns
Orienting T2	−0.10	−1.31	0.03	0.36	0.05	0.86	−0.12	−1.78	8.43*
Child sex	−0.08	−2.54*	−0.08	−2.55*	−0.08	−2.58**	−0.08	−2.58**	ns
Ethnicity	−0.01	−0.26	−0.01	−0.26	−0.01	−0.26	−0.01	−0.26	ns
Income-to-needs ratio	−0.01	−0.57	−0.02	−0.57	−0.02	−0.57	−0.02	−0.57	ns
Orienting T2	R2 = .04**		R2 = .05**		R2 = .04**		R2 = .03**
Orienting T1	0.03	1.05	0.04	1.06	0.03	1.07	0.04	1.07	ns
Cortisol reactivity T1	−0.02	−0.37	−0.02	−0.38	−0.02	−0.38	−0.02	−0.38	ns
Cortisol recovery T1	0.03	0.44	0.02	0.46	0.02	0.45	0.03	0.45	ns
Negative emotion T1	−0.01	−0.32	−0.01	−0.32	−0.01	−0.32	−0.01	−0.32	ns
Child sex	0.05	1.60	0.06	1.58	0.05	1.60	0.05	1.60	ns
Ethnicity	0.19	5.29***	0.23	5.28***	0.20	5.67***	0.16	5.56***	ns
Income-to-needs ratio	−0.01	−0.22	−0.01	−0.22	−0.01	−0.22	−0.01	−0.22	ns
Orienting T3	R2 = .02		R2 = .11		R2 = .05		R2 = .05
Orienting T2	0.09	1.53	0.07	1.54	0.06	1.58	0.08	1.57	ns
Cortisol reactivity T2	0.12	1.70	0.14	1.67	0.09	1.77	0.10	1.73	ns
Cortisol recovery T2	−0.15	−1.49	−0.30	−1.64	−0.03	−0.23	0.08	1.01	12.28**
Negative emotion T2	−0.06	−0.68	−0.25	−3.42***	0.18	2.40*	0.12	1.63	20.07***
Child sex	0.02	0.43	0.02	0.43	0.01	0.43	0.01	0.43	ns
Ethnicity	−0.04	−0.87	−0.04	−0.87	−0.03	−0.88	−0.03	−0.86	ns
Income-to-needs ratio	0.05	1.55	0.05	1.57	0.05	1.58	0.07	1.56	ns

Note.

^*p < .05,

^**p < .01,

^***p < .001.

Child Sex 1= Boy, 2 = Girl. Ethnicity 0 = Black, 1 = Other races. ns = not significant.

The parameter estimates for the multiple-group model are shown in Table 3. Partially supporting our second hypothesis, in all groups, child negative emotion at T2 was associated with higher cortisol reactivity and a greater recovery at T3. Observed negative emotion remained stable over time. Being African American was associated with less cortisol reactivity, more negative emotion, and less orienting behaviors at T2. Girls showed less negative emotion than boys at T3.

We also hypothesized that paths between child orienting behaviors and negative emotion/cortisol responses would be different across parenting profile groups. We found that several paths showed statistical difference among groups (see Table 3 for paths showing significant inter-group differences, and Table 4 for the tests of differences in path coefficients among groups). Orienting to mothers at T1 was associated with increased cortisol reactivity at T2 only in the Intrusive group (β = .13, t = 2.56, p < .01; this association also showed a positive trend in the Engaged group and a negative trend in the Average group). Infants in the Intrusive and Engaged groups had higher cortisol relativity when using orienting behaviors towards their mothers, compared to those in the Average group. The association between cortisol recovery at T2 and child orienting at T3 showed a positive trend in the Engaged group, and a negative trend in the Detached and Intrusive groups, resulting in significant group differences on this association; infants in the Engaged group showed more orienting behaviors after a greater cortisol recovery, compared to those in the Detached and Intrusive groups. Negative emotion at T2 predicted more orienting to mothers at T3 in the Average group (β = .18, t = 2.40, p = .01; significantly different from the Detached and Intrusive groups where this association was negative), but less orienting behaviors in the Intrusive group (β = −.25, t = −3.42, p < .001; significantly different from the Average and Engaged groups where this association was positive). Girls showed less negative emotion at T2 than boys only in the Intrusive group (β = −.25, t = −3.35, p = .001; this association was weaker in all other groups than the Intrusive group).

Table 4.

Differences in Coefficient Estimates among Groups.

Paths	Groups	Mean differences	S.E.	t
Orienting T1 -> Cortisol reactivity T2	Detached - Intrusive	−0.03	0.01	−1.91
	Detached - Average	0.02	0.01	1.74
	Detached - Engaged	0.00	0.01	−0.25
	Intrusive - Average	0.04	0.01	3.29***
	Intrusive - Engaged	0.02	0.01	1.86
	Average - Engaged	−0.02	0.01	−2.04*
Income-to-needs ratio -> Cortisol recovery T2	Detached - Intrusive	0.04	0.03	1.27
	Detached - Average	0.00	0.01	−0.02
	Detached - Engaged	0.00	0.01	0.14
	Intrusive - Average	−0.04	0.03	−1.33
	Intrusive - Engaged	−0.04	0.03	−1.28
	Average - Engaged	0.00	0.01	0.27
Child sex -> Negative emotion T2	Detached - Intrusive	0.87	0.32	2.72**
	Detached - Average	−0.09	0.25	−0.37
	Detached - Engaged	0.22	0.27	0.82
	Intrusive - Average	−0.97	0.29	−3.33***
	Intrusive - Engaged	−0.65	0.30	−2.16*
	Average - Engaged	0.32	0.23	1.37
Orienting T2 -> Negative emotion T3	Detached - Intrusive	−0.14	0.12	−1.16
	Detached - Average	−0.16	0.10	−1.58
	Detached - Engaged	0.01	0.10	0.11
	Intrusive - Average	−0.02	0.11	−0.18
	Intrusive - Engaged	0.15	0.11	1.36
	Average - Engaged	0.17	0.09	1.89
Cortisol recovery T2 -> Orienting T3	Detached - Intrusive	0.37	0.53	0.69
	Detached - Average	−0.47	0.74	−0.63
	Detached - Engaged	−0.96	0.43	−2.24*
	Intrusive - Average	−0.84	0.82	−1.02
	Intrusive - Engaged	−1.32	0.56	−2.39*
	Average - Engaged	−0.49	0.76	−0.64
Negative emotion T2 -> Orienting T3	Detached - Intrusive	0.15	0.10	1.52
	Detached - Average	−0.25	0.12	−2.00*
	Detached - Engaged	−0.16	0.10	−1.64
	Intrusive - Average	−0.40	0.12	−3.26***
	Intrusive - Engaged	−0.30	0.10	−3.19***
	Average - Engaged	0.09	0.12	0.75

Note.

^*p < .05,

^**p < .01,

^***p < .001.

Discussion

The current study investigated the differential patterns of relations among infant negative emotion, behavioral regulation and cortisol responses, associated with different maternal parenting profiles in a low-income sample. The first goal of this study was to identify distinct maternal profiles characterized by parenting behaviors. Four distinct parenting profiles were found, namely Detached, Intrusive, Average and Engaged, consistent with previous observations on maternal parenting behaviors in samples of middle- and upper-classes (e.g., Bowlby, 2012; Field, 2010; Tronick & Reck, 2009). The Engaged group was characterized by high maternal sensitivity, positive regard, animation and stimulation, the Intrusive group by high intrusiveness and negative regard, the Detached group by high detachment, and the Average group by moderate levels of all parenting behaviors. The person-centered approach and findings from observational studies complement each other, and indicate different patterns of parenting behaviors exist in low-income families (Brophy-Herb et al., 2013).

The second goal of this study was to identify the associations between infant negative emotion and cortisol patterns. Consistent patterns were found that an infant’s negative emotion at T2 predicted greater cortisol reactivity and recovery at T3. This indicates that physiological processes are influenced by affective functioning, especially in infancy, during which time these processes are highly malleable (Boyce & Ellis, 2005). Cortisol responses indicate a stable pattern of one’s HPA reactions facing distress, and the changes in cortisol responses usually form over time when exposed to long-term stress. Individual’s affective and behavioral regulation of emotion tend to precede the formation of a consistent cortisol reaction pattern, while providing valuable information about the establishment of such a pattern. Thus, the predictive association between child negative emotion and later cortisol responses suggested the process of infant forming a stable HPA regulation pattern after establishing consistent affective responses (Feldman, 2009; Thompson et al., 2015).

Differential Associations among Aspects of ER across Maternal Parenting Groups

The second goal of this study was to identify different patterns among orienting behaviors, negative emotion and cortisol responses across groups. Infants in the Engaged group showed the highest levels of orienting to mothers at T2, and the positive association between cortisol recovery at T2 and child orienting at T3 was stronger in the Engaged group than the Detached and Intrusive groups. These findings indicate that within a trusting mother-infant relationship, infants turn to their engaged mothers in seeking help for ER, especially after they have had successful experiences in relying on caregivers to regulate emotions (Cassidy, 1994; Diener et al., 2002). Compared to the Average group, infants in the Engaged group showed lower cortisol reactivity at T2, consistent with previous observations that high-quality parenting including sensitive and responsive behaviors reduces infant cortisol reactivity (Ahnert et al., 2004; Albers et al., 2008).

Additionally, among infants of Engaged mothers, the prospective association between orienting to mothers at T1 and cortisol reactivity at T2 showed a positive trend, compared to the Average group where this association showed a negative trend. Similarly, Beijers et al. (2017) revealed that self-soothing strategies were associated with higher concurrent cortisol reactivity, only among securely attached infants. Ekas et al. (2011) observed similar patterns that 20-month-old toddlers using parent-oriented strategies showed increased subsequent negative affect. These findings may seem counter-intuitive. However, Beijers et al. (2017) argued that a higher level of cortisol reactivity may indicate increasing efforts to regulate emotions among securely attached infants, as mothers of these infants can notice their signals of distress and provide assistance. It is also likely that the increased cortisol reactivity among infants of Engaged mothers indicate their engagement in the task, as a result of confidently exploring the environment while using their mothers as a secure base. Future research should investigate these hypotheses.

For infants in the Average group, as expected, negative emotion at T2 was associated with more orienting behaviors at T3, when compared to the Intrusive and Detached groups. This indicates a natural attachment behavior of seeking emotional comfort from their mothers (Bowlby, 2012). Orienting to mothers at T1 was associated with less cortisol reactivity at T2 in the Average group than in the Intrusive and the Engaged groups. This seems to suggest a healthy emotional communication pattern established between infants and their mothers in the Average group. A moderate level of maternal sensitivity, intrusiveness and detachment provides infants with both positive and negative experiences with their mothers. In this situation, negatively reactive infants can freely reach out to their mothers to seek help, and receive assistance for ER (Bowlby, 2012; Cassidy, 1994). As a result, the reduced cortisol reactivity associated with their prior orienting behaviors reveals the process of gradually mastering self-regulation capacities with maternal assistance (Feldman, 2009). Among infants in the engaged group, orienting increased cortisol reactivity which likely prompted future use of ER behaviors; whereas in the average group, orienting reduced cortisol reactivity, and negative emotion at T2 increased future use of orienting. It is likely different qualities of parenting promote children’s use of ER through distinct mechanisms. Sensitive parenting may promote children’s ER through physiological regulation (Hostinar & Gunnar, 2013), whereas children with less sensitive but adequate parenting may achieve this goal through behavioral regulation of affect (Ekas et al., 2011). This hypothesis needs to be tested in future studies.

For infants of Intrusive mothers, orienting behaviors at T1 was associated with higher cortisol reactivity at T2. Although this association was similar to what we observed in the Engaged group, the underlying reason may be different. For infants of intrusive mothers, relaying on orienting to mothers for ER might increase levels of stress in challenging situations, because their mothers often do not provide appropriate support for their infants in time of distress (Ekas et al., 2011; Ispa et al., 2004; Laurent et al., 2016). Negative emotion and cortisol recovery at T2 were associated with less future orienting, as these infants seem to have established an expectation that their mothers may not be sensitive and supportive towards their emotional experiences (Ispa et al., 2004). As a result, infants in this group showed the lowest levels of cortisol recovery at T3 compared to other groups. These findings revealed the negative interaction cycles between intrusive mothers and their children (Cassidy, 1994). Maternal intrusiveness may be a stressful experience for these children and causing these children to experience high levels of physiological responses, while shutting down affective expression (Ahnert et al., 2004; Conradt & Ablow, 2010; Ispa et al., 2004). These children then may be distant from their mothers, and seek less maternal assistance in developing ER skills, and their emotional development can be compromised (Ispa et al., 2004).

For infants in the Detached group, as expected, they showed fewer orienting behaviors towards their mothers at T2 than the Engaged group, the association between cortisol recovery at T2 and child orienting at T3 was negative, comparing to the positive association in the Engaged group. Negative emotion at T2 negatively predicted orienting to mothers at T3 in this group, whereas the same association was positive in the Average group. These associations together characterized the disengaged mother-child relationship in this group; children of detached mothers were less likely to seek maternal support, especially in time of distress (Ekas et al., 2011; Ispa et al., 2004). These infants seem to expect limited assistance from their mothers, thus becoming self-dependent in regulating emotions (Beijers et al., 2017; Diener et al., 2002).

Taken together, findings of this study support the developmental hierarchical-integrative perspective that early physiological, emotional, and behavioral processes of ER are consistently related and impact each other (Feldman, 2009). With the maturation of attentional systems, infants learn to use behavioral strategies such as orienting to facilitate ER development, while integrating physiological, emotional, and behavioral aspects of ER into personalized response patterns. In particular, infants seem to form a stable HPA regulation pattern in the second year of their life after establishing consistent affective responses (Thompson et al., 2015). Further, this study integrated attachment theory, in particular, how infants orient to their mothers for assistance in ER, and how these orienting behaviors initiate and are followed by consistent patterns in affective and physiological responses considering different types of maternal parenting. With sensitive or average parenting, infants using orienting usually leads to adaptive regulation, promoting their future use of the same strategy. Conversely, for infants of intrusive mothers, reliance on orienting behaviors tends to lead to negative outcomes, reducing future use of this strategy. Further, infants with detached mothers avoided using this strategy. Infants’ use of orienting strategy may be based on their expectations of caregiver availability and effectiveness in assisting with their ER, which is crucial in forming early attachment bonds and using mothers as a secure resource for regulation (Bowlby, 2012; Cassidy, 1994). Finally, this study adds to the evidence base of our understanding of infant ER development in low-income families. Mothers in low-income families show similar variations in their parenting behaviors as their peers of a middle- or upper-class background (Ispa et al., 2004; Mesman et al., 2012), whereas these behaviors are associated with different aspects of infant ER development, with implications concerning how to conceptualize the developmental patterns of ER and how to promote intervention efforts in this low-income, at-risk population.

Gender and Racial Differences

Our study revealed early gender differences in ER; girls showed less negative emotion than boys by the end of the second year. This finding is consistent with other literature indicating that infant girls show an earlier ability to regulate emotion than boys (Thompson et al., 2015; Weinberg, Tronick, Cohn, & Olson, 1999). Moreover, girls in the Intrusive group showed less negative emotion at T2 than boys in the same group. This finding is in line with other findings that maternal intrusiveness showed less negative effects on girls than boys (e.g., Eggum et al., 2004; Karreman, Van Tuijl, Van Aken, & Dekovic, 2009). It is likely that mothers are more attentive to highly negative girls than boys, as negative expressions among girls tend to elicit more sympathetic responses from mothers than boys (Chaplin, 2015). On the other hand, boys who are negatively expressive may be especially reactive to mothers’ intrusiveness (Eggum et al., 2004). Our study revealed as early as 15 months, mothers would be responding to male and female infants differently based on gender norms of emotion expression, and these differential responses affect infant’s emotional development.

Additionally, our study revealed the developmental differences in ER across racial groups in this low-income sample. The Engaged group and the Average group included a smaller percentage of African American families than the Detached and Intrusive groups; relatedly, African American infants displayed more negative expressions and less orienting to mothers, which may reflect higher rates of observed insensitive parenting among African American mothers (Mesman et al., 2012). As the quality of parenting is largely a socially constructed concept and can vary across cultures, these parenting behaviors among African American mothers may be closely related to the cultural socialization belief to punish and minimize children’s negative emotion. This cultural socialization belief can be adaptive given African American children can experience greater discrimination when expressing anger (Kang & Chasteen, 2009), and mothers’ encouragement of their children’s negative emotion was related to poorer emotional outcomes for African American, but not White children (Hooper et al., 2018; Nelson et al., 2013). Thus, future researchers should develop more culturally sensitive tools to assess racial differences to capture these culturally adaptive parenting behaviors. Finally, African American infants showed less cortisol reactivity at T2 than White infants, which may be an indicator of an intergenerational adaptation to historical conditions of stress associated with racism and social injustice (Blair et al., 2008).

Strengths and Limitations

Findings of this study should be interpreted with consideration of several limitations. First, this study only included mothers as the primary caregivers, without considering that secure attachment relationships with other caregivers (e.g., fathers) can compensate for the negative effect of an insecure mother-child relationship on infants (Ekas et al., 2011). Future studies may want to consider the joint effect of parenting from multiple caregivers in the family context. Second, child cortisol was only assessed three times at each wave of data collection points. More assessments (5-6 times) would provide a more accurate estimate of cortisol change patterns (Thompson et al., 2015). Third, as this study oversampled low-income, rural families, findings of this study may not be generalizable to moderate- or high-income families. Finally, as emotion socialization is a dynamic, interactive process that maternal and children’s characteristics mutually affect each other, future studies also should consider the change in maternal parenting over time related to the development in children’s behaviors and how it further affects the children reciprocally, to understand how parenting and child behaviors mutually shape each other.

This study also has several strengths, including a high-risk, low-income sample that has been less studied, a longitudinal design to follow infants three times during 1.5 years, and a person-centered approach to identify different maternal parenting profiles, which captures the multidimensional nature of parenting. Second, maternal parenting behaviors and child ER was assessed observationally from laboratory tasks to avoid reporter-biases. Finally, this study identified different patterns between ER and cortisol responses among distinct parenting profiles in a low-income sample, adding to our knowledge of early infant emotional development in a high-risk context. Findings of this study can inform intervention efforts, with a greater understanding of the patterns by which mothers differ and how this impacts child development. Clinicians should tailor interventions according to patterns of familial contextual factors in order to alleviate child risk. For example, a clinician may work with children of intrusive and detached mothers differently to target distinct emotion interaction processes. In all, a more complete comprehension of the ways in which mothers differ in their parenting behaviors and how these variations influence their children will allow for a better understanding in supporting adaptive infant emotional development in the context of poverty.

Data Availability Statement

The data that support the findings of this study are openly available in the Inter-university Consortium for Political and Social Research at https://doi.org/10.3886/ICPSR34602.v4, reference number ICPSR 34602.