Poverty, Caregiving, and HPA-Axis Activity in Early Childhood

Abstract

The association between poverty and the activity of the hypothalamic pituitary adrenal (HPA) axis in early childhood is well established. Both ecological and transactional theories suggest that one way in which poverty may influence children’s HPA-axis activity is through its effects on parents’ behaviors, and over the past three decades a substantial literature has accumulated indicating that variations in these behaviors are associated with individual differences in young children’s HPA-axis activity. More recent research suggests that non-parental caregiving behaviors are associated with HPA-axis activity in early childhood as well. Here we systematically review the literature on the association between both parental and non-parental caregiving behaviors in the context of poverty and the activity of the HPA-axis in early childhood. We conclude by noting commonalities across these two literatures and their implications for future research.

The effects of poverty on child development are a pressing societal concern. Recent estimates suggest that more than 15 million children in the United States are living in poverty (Koball & Jiang, 2018). Decades of research suggest strong linkages between poverty and poor developmental outcomes (see, for example, Evans, Li, & Whipple, 2013; McEwen & McEwen, 2017). These findings have been supported by a growing experimental literature examining income-supplement or anti-poverty programs, all of which suggest that poverty has a negative impact on children’s development (e.g., Costello, Compton, Keeler, & Angold, 2003; Huston et al., 2001; McLoyd, Kaplan, Purtell, & Huston, 2011). These effects may be particularly pronounced for young children (for reviews, see Engle & Black, 2008; Carnegie Task Force, 1994): experiencing poverty in early childhood, as opposed to later childhood, strongly predicts the subsequent development of psychiatric problems (Rutter, 1975), academic difficulties (Lipman, Offord, & Boyle, 1997), and the failure to complete high school (Brooks-Gunn, Guo, & Furstenberg, 1993; Duncan & Brooks-Gunn, 2000; Duncan, Yeung, Brooks-Gunn, & Smith, 1998). Moreover, poverty in early childhood has detrimental effects on measures of attainment 20 to 30 years later, including reduced earnings and work hours (Duncan, Ziol-Guest, & Kalil, 2010).

The strength and persistence of these associations may be explained, in part, by findings from both human and animal research suggesting that there are discrete periods of postnatal development when key neurophysiological processes are consolidated and are especially sensitive to environmental input (Fox, Levitt, & Nelson, 2010; Institute of Medicine & National Research Council, 2015). The effects of early experience on neurophysiological development extend to the organization and consolidation of the stress response system, and one of the main components of that system is the hypothalamic-pituitary-adrenal (HPA) axis (Chrousos & Gold, 1992; Gunnar & Quevedo, 2007). Through the release of corticotrophin releasing factor (CRF), the hypothalamus regulates the release of adrenocorticotropic hormone (ACTH) into the blood stream by the pituitary gland (Gunnar & Vazquez, 2006; Stratakis & Chrousos, 1995). ACTH binds to receptors on the adrenal cortex, promoting the secretion of the glucocorticoid (steroid hormone) cortisol (Gunnar, 1992). Under homeostatic conditions – in which threats or other aversive stressors are absent – cortisol production is relatively low, and much of the cortisol produced is bound and thus rendered inert by mineralcorticoid receptors (MRs; de Kloet, 1991). However, when a threat or stressor is encountered, the production of CRF, ACTH, and cortisol increases. As cortisol levels rise it begins to bind to glucocorticoid receptors (GRs; de Kloet, Rots, & Cools, 1996), which prompts a number of metabolism-enhancing effects (including freeing stores of glucose) that facilitate the behavioral response to challenge (Sapolsky, 1996).

Thus, the term “HPA-axis activity” encompasses at least two distinct types of activity: activity under conditions of homeostasis and activity under conditions of stress or challenge. In older children and adults, the activity of the HPA axis under conditions of homeostasis follows a predictable diurnal rhythm, with levels of cortisol peaking shortly after waking and then declining steadily over the course of the day (Edwards, Evans, Hucklebridge, & Clow, 2001). However, this pattern emerges over the course of early childhood. Newborns and young infants exhibit two peak levels of cortisol approximately 12 hours apart (Klug, Dressendorfer, Strasburger, Kuhl, Reiter, Reich, et al., 2000), and while the circadian rhythm begins to emerge at 3 months (de Weerth et al., 2003; Matagos, Moustogiannis, & Vagenakis, 1998), it is not firmly established until children abandon their naps (Watamura, Sebanc, & Gunnar, 2002). In early childhood HPA-axis activity under conditions of challenge also differs from that seen in older children and adults. Although young infants display elevations in cortisol in response to a variety of stressors (Gunnar, Brodersen, Krueger, & Rigatsu, 1996; Ramsay & Lewis, 1994), between six months and two to three years of age it becomes increasingly difficult to provoke this response (Gunnar et al., 1996; Gunnar et al., 1989; Gunnar & Nelson, 1994; Nachmias, Gunnar, Mangelsdorf, Parritz, & Buss, 1996). Researchers have argued that this dampened stress response may be the human analog of the selective hyporesponsive period of HPA-axis activity observed in young rodents (Gunnar & Quevedo, 2007; Sapolsky et al., 1996).

These age-related changes in HPA-axis activity over the course of early childhood may require defining adaptive and maladaptive HPA-axis activity by referencing patterns of activity that would be considered typical at a given age. For example, failing to exhibit an increase in HPA-axis activity in response to an aversive challenge at three months of age may be maladaptive, given that at three months most children will demonstrate this response (Gunnar et al., 1996; Ramsay & Lewis, 1994). However, at 18 months the same pattern of activity would be considered adaptive, given that a dampened HPA-axis response to challenge is typical at this age (Gunnar & Quevedo, 2007; Sapolsky et al., 1996). Only once patterns of HPA-axis activity have become consolidated is it possible to characterize HPA-axis activity along a single continuum, in which activity at the poles of the continuum may be considered maladaptive and activity towards the center may be considered adaptive.

At one end of this continuum lies hypercortisolism, or frequent and prolonged activation of the HPA axis, which may be manifest as high levels of cortisol under conditions of homeostasis and large increases in cortisol in response to challenge (Fries, Hesse, Hellhammer, &Hellhammer, 2005; Gunnar & Fisher, 2006). Higher levels of cortisol under conditions of homeostasis may be attributable, in part, to “flattened” patterns of diurnal HPA-axis activity, in which the typical decline in cortisol levels over the course of the day is attenuated. These flattened patterns of cortisol are often accompanied by lower levels of waking cortisol, with the net result being elevated overall cortisol production throughout the day (Miller, Chen, & Zhou, 2007). At the other end of the continuum is hypocortisolism, which is indicated by very low levels of cortisol under homeostatic conditions and a “blunted” response to challenge (Fries, Hesse, Hellhammer, & Hellhammer, 2005; Gunnar & Fisher, 2006). These altered patterns of HPA-axis activity may constitute a recalibration or allostasis in response to environmental conditions (McEwen & Seeman, 1999). Specifically, exposure to acute stress that is extreme in intensity or duration may result in hypercortisolism, whereas hypocortisolism may be indicative of habituation of the HPA axis to chronic stress (Fries, Hesse, Hellhammer, & Hellhammer, 2005; Gunnar & Fisher, 2006). Hypercortisolism may, in fact, lead to hypocortisolism over the course of development, as the hippocampal feedback circuits that modulate HPA-axis activity become dysregulated by exposure to elevated levels of cortisol (Gunnar & Vasquez, 2001; Susman, 2006). Although these patterns of altered HPA-axis activity may be adaptive in the short term, over time they may impose a heavy allostatic load on children’s development (McEwen, 1998; McEwen, 2000). For example, hypercortisolism may increase the risk for cardiovascular disease and insulin resistance, impair immune functioning, and promote atrophy in the hippocampus and prefrontal cortex (Bremner & Vermetten, 2001; McEwen, 2006), whereas hypocortisolism has been linked with autoimmune and chronic pain disorders, disrupted self-regulatory abilities, and inhibited growth (e.g., Blair, Granger, & Razza, 2005; Gunnar & Vazquez, 2001; Heim, Ehlert, & Hellhammer, 2000).

The malleability of HPA-axis function may explain how the experience of poverty gets “under the skin” of young children (Lupien et al., 2001). The environment of poverty is filled with stressors for children and their caregivers. Families in poverty are more likely to live in homes with holes in the floor and ceiling, inadequate heat and plumbing, rodent or other pest infestations, greater crowding, noise pollution, and toxins (for reviews, see Children’s Defense Fund, 1995; Evans, 2004), and in concentrated areas of poverty there may be increased rates of criminal activity, violence, and homicides (Sampson, Raudenbush, & Earls, 1997). Boyce and Ellis (2005) argue that early exposure to environmental adversity promotes hyperactivity of the physiological response to stress, and, in accordance with this account, studies employing variable-centered (Lupien et al., 2001; Zalewski et al., 2012a) and person-centered (Holochwost et al., 2017; Laurent et al., 2014) approaches have linked poverty to dysregulated HPA-axis activity in young children. Dysregulation of the HPA-axis activity in early childhood has, in turn, been found to predict subsequent outcomes across a number of domains, including social and cognitive function (e.g., Blair et al., 2005).

In this paper we explore what imbues poverty with the power to influence development, and, in particular, the HPA axis. While there is no single answer to such a complex question, an integration of ecological (Bronfenbrenner & Morris, 1998) and transactional (McLoyd, 1998) theories suggests that poverty – an aspect of the child’s distal environment – affects development through its influence on more proximal systems. For the young child, relationships with parental caregivers are a particularly salient aspect of the proximal environment, and thus the tendency for poverty to erode parental caregivers’ capacity to engage in positive behaviors while increasing the likelihood that they will engage in negative behaviors (Holochwost et al., 2016; Popp, Spinrad, & Smith, 2008), may, in part, explain how poverty influences HPA-axis activity (Institute of Medicine & National Research Council, 2015). And yet this account is incomplete, given that the social conditioning of HPA-axis activity occurs through interactions with caregivers other than children’s parents (Flinn, 1999; Gunnar & Quevedo, 2007; Hostinar, Sullivan, & Gunnar, 2014) and that many young children spend a considerable amount of their time with non-parental caregivers in early care and education settings (Capizzano & Main, 2005; Corcoran & Steinley, 2017). Therefore, and in accordance with the contextual systems model (CSM; Pianta & Walsh, 1996), in order to understand how poverty may influence HPA-axis activity through its effects on parental caregiving behaviors, it is necessary to understand the influence that poverty may exert through non-parental caregiving behaviors as well.

Here we present a systematic review of the literature on the associations between multiple dimensions of parental and non-parental caregiving behaviors and different aspects of young children’s HPA-axis activity. By reviewing studies that examined parental and non-parental caregiving behaviors, it was possible to identify points of convergence and divergence in findings across the parenting and early care and education literatures. By necessity, our review focuses on the direct associations between caregiving behaviors and HPA-axis activity, given that few studies have assessed the question of mediation. Although the results of our review clearly support the plausibility of a model in which parental and non-parental caregiving behaviors mediate the association between poverty and HPA-axis activity, this does not render an alternative model in which these behaviors moderate this association implausible. In the discussion we outline specific directions for future research designed to test the mediation model that is the focus of this paper, as well as considerations that may complicate this model.

Parental Caregiving Behaviors in the Context of Poverty

As noted above, families in poverty often face dangerous living situations in their homes and neighborhoods (Children’s Defense Fund, 1995; Evans, 2004), but impoverished parents face a host of additional stressors beyond the physical environment. For example, the employment opportunities available to low-income parents are often unstable and high in stress, involving non-standard shifts, long work hours, and little self-direction (Hsueh & Yoshikawa, 2007; Parcel & Menaghan, 1997). Impoverished families may also experience greater social isolation than those with more financial resources (Belle, 1983), and their family and friend networks may add additional strain due to their own financial difficulties (Brodsky, 1999). The pressure of being forced to make cutbacks to make ends meet or the inability to provide for family necessities can add further tension to parents’ lives (e.g., Conger, Wallace, Sun, Simons, McLoyd, & Brody, 2002).

In extreme cases, poverty may increase the likelihood of child maltreatment (Drake & Zuravin, 1998). A disproportionate number of families in poverty are reported for abuse (Trickett, Aber, Carlson, & Cicchetti, 1991), and the sensations of powerlessness that living in poverty can create may further heighten the potential for maltreatment (Bugental & Happaney, 2004). Maltreatment in early childhood has, in turn, been linked to altered patterns of HPA-axis activity under conditions of both homeostasis (Bernard, Butzin-Dozier, Rittenhouse, & Dozier, 2010; Bruce, Fisher, Pears, Levine, 2009) and challenge (Bugental, Martorell, Barraza, 2003; Hart, Gunnar, & Cicchetti, 1995).

However, for the vast majority of families in poverty, parenting behaviors do not encompass these extremes. Nevertheless, the experience of poverty may reduce parents’ ability to be sensitive and supportive while increasing the likelihood that they will engage in sub-optimal parenting behaviors (Bradley & Corwyn, 2001; McLoyd, 1990). For example, during interactions with their infants, impoverished mothers are more likely to be rated as insensitive than more affluent mothers (Pianta & Egeland, 1990; Shaw & Vondra, 1995) and to exhibit lower levels of emotional responsiveness (Klebanov, Brooks-Gunn, & Duncan, 1994). Parents in poverty are less likely to consider their children’s perspectives and empathize with their interests (Gerris, Deković,& Janssens, 1997), which may lead to higher levels of parent-child conflict over time (Hastings & Grusec, 1997).

Non-Parental Caregiving Behaviors in the Context of Poverty

Although parental caregiving behaviors are the sine qua non among influences on children’s development, many young children spend a substantial proportion of their waking hours in the care of people other than their parents. Recent data from the Early Childhood Program Participation (ECPP) Survey indicated that nationally 60% of children age 5 years or younger were in at least one form of non-parental care (Corcoran & Steinley, 2017); analyses of earlier data from the National Survey of America’s Families (NSAF) found that 42% of children under 5 spend at least 35 hours a week in care (Capizzano & Main, 2005). Despite the presence of barriers to access of affordable care (see below), many children in poverty do spend time in care: according to the ECCP Survey, 46% of these children are in non-parental care (Corcoran & Steinley, 2017), while NSAF data indicated that for children in low-income, single-parent households this figure was 79% (Adams et al., 2007). Indeed, welfare reform (the Personal Responsibility and Work Opportunity Act of 1996) required single mothers of infants to work, effectively mandating that many low-income children be in non-parental care (National Research Council and Institute of Medicine, 2000).

While the non-parental caregiving arrangements for children in poverty vary widely (Zaslow et al., 2006), there is evidence that, on average, the quality of care is lower for poor children than it is for their more affluent peers (Adams et al., 2007; Howes & Olenick, 1986; NICHD Early Child Care Research Network, 1997; Phillips, Voran, Kisker, Howes, & Whitebook, 1994). Despite the importance of consistent caregiving arrangements for children’s development (Institute of Medicine and National Research Council, 2015; Rhodes & Huston, 2012; Tout et al., 2005), rates of turnover among early educators are among the highest of any profession (National Research Council and Institute of Medicine, 2000; Whitebrook et al., 2014), and these rates may be higher among caregivers of children in poverty, given the low wages paid to many of these caregivers (Institute of Medicine and National Academy of Sciences, 2015; Whitebrook et al., 2014). Over a third of children in poverty (37%) are in multiple caregiving arrangements (Zaslow et al., 2006), a figure that excludes the common practice of cobbling together care on a day-to-day basis (Adams et al., 2007).

Even when care is stable, children in poverty are more likely to attend home-based care (NSECE Project Team, 2016), which may include relative care (Corcoran & Steinley, 2017). These types of care arrangements are generally less expensive and offer more flexibility than center-based care for parents who work long, irregular, or overnight hours (Hofferth, 1995; Porter, Paulsell, Nichols, Begnoche, & Del Grosso, 2010). However, they may also be of lower quality than center-based alternatives (Brown-Lyons, Robertson, & Layzer, 2001; Galinsky, Howes, Kontos, & Shinn, 1994; Tout & Zaslow, 2006), given the lower levels of resources and supports offered to providers of home-based care (Tonyan, Paulsell, & Shivers, 2017) and the fact that some forms of home-based care (e.g., family, friend, and neighbor care) are exempt from licensing requirements (Zaslow, Tout, & Martinez-Beck, 2010). When center-based care is available for children in poverty, it is likely to be of lower quality than center-based options for more affluent children (Marshall et al., 2001). For example, whereas ratings of Head Start classrooms have consistently reported quality scores at the threshold between “low” or “minimal” and “good” quality using a combination of measures (e.g., the Early Childhood Environmental Rating Scale (ECERS; Harms & Clifford, 1980) and the Classroom Assessment Scoring System (CLASS; Pianta, La Para, & Hamre, 2008); Aikens, Bush, Gleason, Malone, & Tarullo, 2016), average levels of quality are lower for classrooms that serve a majority of low-income children (Pianta et al., 2005).

There are a variety reasons that children in poverty are apt to receive lower-quality care, many of them market-driven (Adams et al., 2007). The structural features of high-quality care, such as specialized training for caregivers and low child-to-caregiver ratios, are expensive to provide (Marshall et al., 2001), and when the cost of this care is not subsidized it may be beyond the means of many low-income families (National Research Council & Institute of Medicine, 2000). But these structural aspects of quality are a prerequisite, rather than a guarantor, of high-quality care. As noted by the National Research Council and Institute of Medicine (2000):

Quality of care ultimately boils down to the quality of the relationship between the child care provider or teacher and the child. A beautiful space and an elaborate curriculum – like a beautiful home – can be impressive, but without skilled and stable child care providers, they will not promote positive development…Critical to sustaining high-quality child care for young children are the providers’ characteristics, notably their education, specialized training, and attitudes about their work and the children in their care, and the features of child care that enable them to excel in their work and remain in their jobs. (pp. 314–315, 318).

This account makes clear why non-parental caregivers of children in poverty may struggle to establish and maintain positive interactions with the children in their care. Like many parents of children in poverty, these caregivers are engaged in a stressful, demanding job (Friedman-Krauss, Raver, Neuspiel, & Kinsel, 2014), often with limited professional support (Institute of Medicine and National Academy of Sciences, 2015), and, like many impoverished parents, caregivers of poor children are more likely to face social isolation (Porter et al., 2010), elevated levels of financial stress (Whitebrook et al., 2014) and depression (Whitaker et al., 2013; Whitebrook & Sakai, 2004). Indeed, many caregivers of children in poverty are themselves poor. On average, childcare workers earned $20,660 per year in 2013, placing them just above the poverty level for the family of three (Bureau of Labor Statistics, 2014). Caregivers in home-based programs earned slightly less ($19,510 per year; Bureau of Labor Statistics, 2014). It is perhaps not surprising that between 35 and 46% of caregivers reported using one or more forms of public assistance (Whitebrook et al., 2014). In short, many of the aspects of sociodemographic risk that have been found to erode parents’ ability to interact positively with their children – poverty, single-parenting status, and low levels of education – are present in large proportions of the caregiving workforce, of whom 16% fall below the poverty line, 17% are single parents, and 50% hold no more than a high school diploma (Whitebrook et al., 2014). Indeed, as Whitebrook and colleagues (2014) noted, many caregivers of children in poverty are also parents of children in poverty.

The Current Study

In summary, the activity of the HPA-axis is open to environmental input in early childhood, including input from environments of poverty and adversity (Holochwost et al., 2017; Laurent et al., 2014; Lupien et al., 2001; Zalewski et al., 2012a). Accordingly, and consistent with ecological systems (Bronfenbrenner & Morris, 1998) and transactional theories (McLoyd, 1998), caregiving behaviors may transmit the effects of poverty to young children’s HPA-axis activity. Moreover, according to the contextual systems model (Pianta & Walsh, 1996) it is not possible to fully understand the association between one form of caregiving behavior, whether parental or non-parental, and HPA-axis activity without understanding the parallel association between the other form of caregiving behavior and HPA-axis activity. Thus, the fact that associations between various dimensions of parental caregiving behaviors and different aspects of HPA-axis activity are often reported without reference to parallel associations between non-parental caregiving behaviors and HPA-axis activity has impeded a fuller understanding of an essential mechanism by which poverty may influence the activity of the HPA axis in young children. So too has the tendency to report associations between non-parental caregiving behaviors and HPA-axis activity without acknowledging findings from the parenting literature. Therefore, in the remainder of this paper we present the results of a systematic review of the literature on the association between caregiving behaviors, first reviewing studies of parental caregiving behaviors in the context of poverty and then studies of non-parental caregiving behaviors in the same context. To the extent possible, the results of these studies are presented in a similar fashion to facilitate the identification of points of convergence and divergence between these two literatures.

Method

Search Strategy

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were used to develop a protocol for reviewing the literature (Shamseer et al., 2015). To be included in the review, studies had to meet the following inclusion criteria: 1) participants were young children (birth to age 8 years) who were healthy and developing normally and 2) their caregivers, whether parental or non-parental; 3) some dimension of parental or non-parental caregiving behaviors were assessed; 4) some aspect of HPA-axis activity was measured, and 5) the study was designed to examine the association between some dimension(s) of caregiving behavior and some aspect(s) of HPA-axis activity. In addition to these study inclusion criteria, a series of report characteristics were also applied as inclusion criteria. Therefore, to be included in the review, a study also had to be: 1) empirical (which excluded letters, commentary, and case studies); 2) conducted with human children who were neonates, infants, or preschoolers; and 3) were published in English in a peer-reviewed journal. No date limitations were placed on the search results. Previous research has demonstrated that excluding papers published in languages other than English does not necessarily bias the results of reviews (Morrison et al., 2012), and including only papers that underwent peer review is one means of ensuring the quality of studies (cf., Lawson et al., 2018) without resorting to potentially-problematic quality assessment tools (Siddaway et al., 2019)

Two databases (PsycInfo and Medline) were searched concurrently using the following combination of keywords that were based on the study inclusion criteria: [“parent*” OR “childcare” OR “caregiv*” OR “teacher*” OR “early education” OR “preschool”] AND [“cortisol” OR “hypothalamic*” OR “HPA*”]. The results of the search were then narrowed by applying the reporting inclusion criteria. The reference lists of the studies that were ultimately included in the review following full-text screening (see below) were also consulted to identify any additional records not yielded through the database search. The final search was conducted in June, 2019.

Study Selection

The records yielded by the database search, including titles and abstracts, were exported into a spreadsheet. The title and abstract of each study was screened by the first author using the study inclusion criteria. When the decision was made to exclude a study, the reason for doing so was selected from one of the following options: 1) participants were not young children (defined as birth to age 8 years) and 2) their caregivers; 3) no caregiving behaviors were assessed; 4) no aspect of HPA-axis activity was measured, or 5) the study did not examine the association between some dimension(s) of caregiving behavior and some aspect(s) of HPA-axis activity.

These same inclusion criteria were then applied to the full texts of the papers that passed the review of titles and abstracts. The full texts of all studies were reviewed by the first author. Approximately half (51%) of these full texts were selected at random for independent review by either the fourth or final author, with both of these authors assigned approximately half of the full texts selected for independent review. When the decision was made to exclude a study, the reviewing author made a note of the reason using one of the options listed above. The first and second readers agreed in each case about which papers to include in the review, and, in those cases when the decision was made to exclude a study, both readers agreed about the reason for doing so. Finally, the first author applied the inclusion to the reference lists of the studies that passed the full text review.

Data Extraction and Analysis

Once the studies that were to be included in the review had been identified, the fourth author extracted relevant data using a form that included the following fields: 1) type(s) of caregiving behaviors (parental/non-parental); 2) sample size; 3) the composition of the sample in terms of age and socioeconomic composition; 4) the dimension(s) of caregiving behaviors assessed and how those dimensions were measured; 5) the aspect(s) of HPA-axis activity measured and the method and timing of measurement; and 6) the major findings of the study. This form was developed and piloted by the first and final authors, who then together trained the fourth author in its use. As data were extracted by the fourth author, the first author reviewed all entries in the form and asked clarifying questions as appropriate to ensure reliability in the data extraction process. During the extraction process two points became clear: first, it was necessary to add a notes section to the extraction form about important design issues, such as whether a study was a randomized-control trial of an intervention for caregivers. Second, a number of studies used the same or partially-overlapping samples, and therefore this was noted on the form as well. Given that these studies examined different associations between dimensions of caregiver behaviors and aspects of HPA-axis activity and that the results of the review were not subject to quantitative analysis (see below), studies that used the same were retained.

The data from the extraction form were reformatted as a pair of tables that reflected the goal of the study: to systematically review the literature on the associations between dimensions of parental and non-parental caregiving behaviors and different aspects of HPA-axis activity in early childhood. As such, Table 1 included the data listed above that was drawn from studies of parental caregiving behaviors, while Table 2 presented the analogous data drawn from studies of non-parental caregiving behaviors. Where multiple studies were based on the same sample this is noted in the table and text. Both tables reported the same extracted data, and these data were reported in a way that facilitated the comparison of findings across studies of parental and non-parental caregiving. For example, diurnal HPA-axis activity was defined as activity observed under homeostatic conditions at multiple timepoints over the course of the day, and the timing of data collection was reported for all studies that examined the association between caregiving behaviors, whether parental or non-parental, and diurnal HPA-axis activity, thus defined.

Table 1.

Studies of parental caregiving behaviors and HPA axis activity

Study	Participants		Assessment of Parenting Behaviors		Measurement of HPA Axis Activity		Summary of Results
	N	Age & Socioeconomic Composition	Construct(s)	Measurement	Conditions	Method & Timing	Major Findings	Notes
Albers et al., 2008	64	Infants & Toddlers (M=12.2 mos.) (SD=2.5 mos.) Income not reported. Mean maternal education was 6.22 on a scale ranging from 1 (elementary school) to 7 (university degree).	Maternal sensitivity and cooperation	Observation during bath	Baseline, Reactivity (bath)	Salivary cortisol was obtained before bath and then 25 and 40 m after bathing.	Higher levels of maternal sensitivity and cooperation during bath time were associated with faster recovery of cortisol levels to baseline levels.
Albers et al., 20161	64	Infants (M=14.6 wks.) (SD=2.8 wks.) Income not reported. 34 mothers (53.1%) had a professional degree and 29 (45.3%) had at least one university degree.	Maternal sensitivity and cooperation	Observation during bath	Diurnal	Salivary cortisol was obtained in the morning (10–10:30 a.m.) and afternoon (4–4:30 p.m.) while children were in care.	Higher levels of maternal sensitivity and cooperation during bath time exhibited higher morning cortisol while in care. Similar results were observed in the afternoon only among infants rated high in negative emotionality.
Bakermans-Kranenburg et al., 2008	130	Toddlers (M=28.3 mos.) (SD=10.3 mos.) Income not reported. Mean maternal education was 3.72 on a scale ranging from 1 (elementary school) to 5 (university degree).	Maternal sensitivity and discipline	n/a	Diurnal	Salivary cortisol was collected at home by parents on a typical day within a week before the post test upon children’s waking, before lunch, and in the afternoon.	Receipt of an intervention (VIPP-SD2) designed to improve maternal behavior was associated with reductions in diurnal cortisol levels among children with the DRD4 7-repeat allele.	Intervention study (featuring random assignment to a program designed to improve parental caregiving behaviors).
Bernard & Dozier, 2010	32	Toddlers (M=15.2 mos.) (SD=2.3 mos.) Annual household income ranged from $10,000 to $100,000, with 53% of households earning more than $100,000.	Attachment security and organization	Observation during free play and the SSP3	Baseline Reactivity (SSP)	Salivary cortisol at at the child’s home, upon the child’s arrival at the lab, and 40, 65, and 80 m after arrival.	Infants rated as disorganized exhibited elevations in cortisol in response to free play and larger elevations in response to the SSP; infants rated as organized infants did not exhibit elevations in response to either task.
Berry et al., 20174	1292	Infants & Toddlers (7–24 mos.) The majority (67%) of families had an annual household income of less than 200% of the federal poverty level.	Maternal sensitivity	Observation during a semi-structured interaction session5	Baseline	Salivary cortisol was collected at home at 7, 15, and 24 mos. of child age.	Children whose mothers displayed lower levels of sensitivity across ages exhibited higher levels of overall, cross-age cortisol. Among mother-child dyads characterized by low overall levels of maternal sensitivity, increases in maternal sensitivity over time were associated with decreases in cortisol levels. Among mother-child dyads characterized by high overall sensitivity, increases in sensitivity were associated with increases in cortisol levels.	Analyzed between-child differences in levels of cortisol across age and longitudinal, within-child associations between sensitivity and cortisol.
Blair et al, 2015	1292	Toddlers (M=24 mos.) The majority (67%) of families had an annual household income of less than 200% of the federal poverty level.	Parental sensitivity (over 90% mothers)	Observation during a semi-structured interaction session3	Baseline Reactivity (a fear-inducing mask presentation following a toy removal procedure6)	Salivary cortisol was collected prior to the stress task, ∼20 m after the child reached peak emotional arousal, and ∼40 m after peak arousal.	Children evidencing emotional liability to the fear-inducing mask and whose parents displayed higher levels of sensitivity exhibited elevated baseline cortisol and greater reactivity. The association between liability and cortisol was not observed among children whose parents displayed lower levels of sensitivity.	The focal predictor of cortisol was children’s emotional liability; parental sensitivity was included as a moderator.
Blair et al., 2008	1292	Infants (M=7.6 mos.) Toddlers (M=15.7 mos.) The majority (67%) of families had an annual household income of less than 200% of the federal poverty level.	Maternal positive engagement and intrusiveness	Observation during free play	Baseline, Reactivity (Infancy7) (Toddlerhood8)	Salivary cortisol was collected prior to free play, 20 m after peak emotional arousal, and 40 m after peak arousal.	Maternal engagement in infancy was associated with: 1) higher levels of reactivity in infancy, and 2) lower overall levels of cortisol in toddlerhood.	Peak arousal was determined by the data collectors using clear guidelines established in the experimental protocol. Peak arousal for the majority of infants occurred at the conclusion of the emotional challenge tasks at both time points.
Blair, Granger, et al., 2011	1292	Infants (∼7 mos.) Toddlers (∼15 & 24 mos.) Preschoolers (∼36 mos.) The majority (67%) of families had an annual household income of less than 200% of the federal poverty level.	Maternal positive parenting9 and negative10 parenting	Observation during free play (7 and 15 mos.)	Baseline	Salivary cortisol was collected prior to free play.	Baseline cortisol levels were negatively related to positive parenting across age, but unrelated to negative parenting.	In combination with positive and negative parenting and household risk, cortisol mediated effects of income-to-need, maternal education, and African American ethnicity on child cognitive ability.
Blair, Raver, et al., 2011	1135	Infants (∼7 mos.) Toddlers (∼15 & 24 mos.) Preschoolers (∼36 & 48 mos.) The majority (67%) of families had an annual household income of less than 200% of the federal poverty level.	Maternal positive parenting and negative parenting	Observation during free play (7 and 15 mos.)	Baseline	Salivary cortisol was collected prior to free play.	Lower levels of maternal sensitivity were associated with higher levels of baseline cortisol from 7 to 48 mos.
Bugental et al., 2003	44	Toddlers (M=17.6 mos.) (SD=4.7 mos.) Income not reported. Participating mothers were classified as high risk on the basis of low education, low social support, high stress, and a history of abuse as children.	Maternal emotional withdrawal11	Self-repot (CTS)12	Baseline Reactivity (SSP13)	Salivary cortisol was collected after an unstructured interaction between mothers and infants and 20 m after the SSP.	Frequent maternal emotional withdrawal was associated with elevated baseline cortisol levels.
Bugental et al., 2010	64	Infants (M=9 wks.) (SD=5.5 wks.) Income not reported. Participating mothers were classified as high risk on the basis of low education, residential instability, and limited English language proficiency.	Parenting knowledge	Self-report (CTS)	Basal	Salivary cortisol was collected in the morning (∼10 a.m.) once a year over the course of three years.	At the 1- and 3-year assessments, infants of parents assigned to the treatment condition exhibited lower levels of baseline cortisol. Infants of parents assigned to the treatment condition also exhibited declining cortisol over time.	Intervention study (featuring random assignment to a year-long program designed to enhance caregivers’ knowledge about parenting).
Cicchetti et al., 2011	143	Infants & Toddlers (M=13.3 mos.) (SD=0.8 mos.) Mean annual household income was $16,646. 96% of participating families were receiving public assistance.	Maternal sensitivity	History of maltreatment served as a proxy measure	Basal	Salivary cortisol was collected in the morning four times over the course of two years.	Maltreated infants whose families received standard services exhibited declining cortisol levels over time. Maltreated infants whose families participated in the program exhibited cortisol levels that were no different from those exhibited by children who were not maltreated	Intervention study (featuring random assignment to a year-long attachment-based program designed to enhance parental caregivers’ sensitivity).
DePasquale et al., 2018	66 children from international orphanages or foster care14	Toddlers (M=25.0 mos.) (SD=4.6 mos.) All adoptive families reported annual household incomes of at least $40,000, with 5% reporting incomes between $40,000 and $59,000, 32% between $60,000 and $99,000, and the remainder over $100,000.	Matemal sensitivity	Observation during free play	Baseline Reactivity (SSP)	Salivary cortisol was collected immediately before and then 15 m and 30 m after the SSP.	Children whose adoptive mothers displayed higher levels of sensitivity exhibited larger increases in cortisol in response to the SSP.	Maternal sensitivity was not associated with baseline cortisol levels.
Dougherty et al., 2011	160	Preschoolers (M=43.5 mos.) (SD=2.8 mos.) Income not reported. The majority of children were from two-parent, middle-class families.	Parental hostility (96% mothers)	Observation during Teaching Tasks battery15	Baseline Reactivity (Lab-TAB16)	Salivary cortisol was collected at baseline and after a Stranger Approach task (60 m post-baseline), a Transparent Box task (90 m), and a Box Empty task (130 m).	Parental hostility was associated with high and increasing cortisol levels among children whose parents had a history of depression.	This moderating effect was specific to children who were exposed to maternal depression during the first few years of life.
Dougherty et al., 2013	153	Preschoolers (M=49.9 mos.) (SD=9.8 wks.) Annual household income ranged from less than $20,000 (7% of the sample) to over $100,000 (36%).	Maternal hostility and support	Observation during five tasks (a book reading, a guessing game, a maze, a story sequencing task, and a set of puzzles)	Baseline Reactivity (Matching task17)	Salivary cortisol was collected prior to the stressor task (after a 30-m quiet play session) and 20, 30, 40, and 50 m following the completion of the matching task.	Maternal hostility moderated the association between maternal depression and reactivity. Children of mothers who had a history of depression during the child’s life and who displayed hostility exhibited increases in cortisol in response to the stressor, while children of mothers without depressive history who displayed hostility exhibited decreases in cortisol.
Feldman et al., 2010	53 mother-infant dyads (randomly assigned to SFP or SFP+T)18	Infants (M=26 wks.) Income not reported. Mothers had at least a high school education.	Maternal touch	Observation during SFP or SFP+T	Baseline Reactivity (SFP or SFP+T)	Baseline salivary cortisol was collected upon arrival at the lab in the morning. Cortisol samples were collected 20 and 35 m after the SFP or SFP+T procedure	Cortisol reactivity was higher among infants in the SFP condition relative to those in SFP+T condition. Cortisol decreased at recovery for infants in the SFP+T, but further increased for those in the SFP.	Myssynchrony was also measured, and indicated the frequencies of maternal stimulatory and proprioceptive touch while the child showed gaze aversion. Myssynchrony was associated with higher cortisol levels among children.
Fisher, Serbin, et al., 2007	36	Preschoolers (M=4.8 yrs.) (SD=1.1 yrs.) Median annual household income was $31,725. Fourteen children (35%) came from families with incomes below the Canadian threshold for poverty.	Maternal emotional support and stimulation	Observation during a 15m task19	Diurnal	Salivary cortisol samples were collected by mothers upon child wakening, ∼20 m post waking, and again every two hours until child went to sleep at night.	Children of mothers who displayed lower levels of maternal emotional support exhibited lower levels of cortisol ∼20 m post waking.
Fisher, Stoolmiller et al., 2007	177	3 groups of children20 (M=4.4 yrs.) Median annual household income was between $15,000 and $19,999.	Maternal consistency and responsiveness	n/a	Diurnal	Early morning and evening salivary cortisol levels were assessed over 12 mos., twice per month. Morning: 30 m after waking Evening: 30 m before bedtime	Children of foster parents receiving the intervention exhibited patterns of diurnal cortisol comparable to those observed among non-maltreated children by the end of the study. Children of foster parents not receiving the intervention exhibited increasingly ‘flattened’ patterns of diurnal cortisol by the end of the study.	Intervention study (featuring random assignment to a program designed to teach consistent and responsive parental caregiving behaviors).
Fisher et al., 2011	71	Preschoolers (M=4.47 yrs.) Median annual household income was between $15,000 and $19,999.	Maternal consistency and responsiveness	n/a	Diurnal	Saliva cortisol was collected in the morning and evening on 2 consecutive days each month. Initial collection occurred 3–5 wks. after a placement change Morning: 30 m after waking Evening: 30 m before bedtime	Placement change was associated with increasingly smaller diurnal declines in cortisol (i.e., higher diurnal cortisol levels) in the regular foster care group. In the intervention foster care group, elevations in diurnal cortisol were not observed.	Intervention study (featuring random assignment to a program designed to teach consistent and responsive parental caregiving behaviors).
Grant et al., 2009	88	Infants (M=37 wks.) (SD=0.76 wks.) Income not reported. Participating mothers were classified as high risk on the basis of a score of 23 or higher on the antenatal risk questionnaire (ANRQ).21	Maternal sensitivity	Observation during free play	Baseline Reactivity (SFP)	Salivary cortisol was collected upon arrival at the lab, and 15, 25, and 40 m after the still-face procedure.	Lower levels of maternal sensitivity were associated with higher levels of reactivity.
Gunnar et al., 1996	73	Toddlers (M=18 mos.) Income not reported. Most (85%) mothers held bachelors degrees or higher.	Maternal responsiveness & attachment security	Observation during the well-baby exam and inoculation.	Baseline Reactivity (Inoculation, SSP)	Salivary cortisol was collected 20 m after inoculation. For SSP, salivary cortisol was collected upon arrival at the lab and 25 m following the first separation (10 m following the last reunion).	At 15 and 18 mos., insecurely attached infants who were also classified as being high on fearfulness were found to display higher cortisol reactivity to inoculation and the SSP At 2 to 6 mos., baseline cortisol levels were significantly lower for securely attached babies.
Hastings et al., 2011	402	Preschoolers (M=4.01 yrs.) (SD=0.71 yrs.) Income not reported. Highest level of maternal education ranged from high school (22%) to an advanced degree (40%).	Maternal punishment	Self-report22	Baseline Reactivity (interactions with unfamiliar adults)	Salivary cortisol was collected ∼20 m after meeting the unfamiliar adults, after interactions with the adults, and again ∼40–60 m after the first sample.	Children of mothers who reported engaging in higher levels of punishment exhibited larger increases in cortisol in response to interactions with unfamiliar adults.
Hertsgaard et al., 1995	38	Toddlers (M=19.3 mos.) Income not reported.	Attachment security and organization	Observation during free play and the SSP23	Reactivity (SSP)	Salivary cortisol was collected 30 m after the beginning of first separation episode.	Children classified as disorganized in their attachment exhibited higher levels of cortisol in response to the SSP.	Baseline cortisol levels were not assessed.
Hutt et al., 2013	9366	Toddlers (M=24 mos.) Annual household incomes ranged from below $30,000 (7% of the sample) to above $60,000 (50%).	Maternal protective behavior	Observation during a series of novel tasks24.	Baseline Reactivity (Novel tasks)	Salivary cortisol was collected prior to (∼11:30 a.m.) and following (∼1:00 p.m.) the completion of the novel tasks	Higher levels of maternal protective behavior were associated with larger increases in cortisol in response to the novel tasks.
Jansen et al., 2010	141	Infants (M=34.5 days.) (SD=3.7 days.) Income not reported. Most (75%) participants were rated as having a high level of education (not defined).	Maternal sensitivity and cooperation	Observation during bath	Baseline Reactivity (Bath)	Salivary cortisol was collected within 10 m of researcher’s arrival at the participants’ home, and 25m and 40m after the infant was taken out of the bath.	The quality of maternal caregiving was not associated with either cortisol reactivity or recovery.
Johnson et al., 2018	177	Toddlers (M=15.3 mos.) (SD=2.61 mos.) Over one third (35%) of participants had annual household incomes under 100% of the federal poverty level; an additional 17% had incomes between 100% and 150% of the federal poverty level.	Attachment security	Observation during child clinic visit (Q-Sort25)	Baseline Reactivity (Inoculation)	Salivary cortisol was collected upon arrival at the clinic arrival, immediately following the physical exam before the inoculation, and 20 m post inoculation.	Toddlers rated as insecurely attached and who were from households below 150% of the federal poverty level exhibited significantly higher baseline cortisol than toddlers who rated as securely attached from comparably-impoverished households. Reactivity in response to inoculation did not differ by poverty or attachment security.
Kaplan et al., 2008	47	Infants (M=4 mos.) Income not reported. Nearly all (93%) mothers had a high school diploma or higher.	Maternal sensitivity	Observation during free play	Baseline	Salivary cortisol was collected before the free play session between.	Maternal sensitivity moderated the effects of antenatal psychiatric illness on infant baseline cortisol levels. Infants of mothers with a psychiatric diagnosis who displayed low maternal sensitivity exhibited higher baseline cortisol levels.	Maternal sensitivity was unrelated to cortisol among infants of mothers without a psychiatric diagnosis.
Kertes et al., 2009	274	Preschoolers (M=3.97 yrs.) (SD= 0.48 yrs.) Annual household income ranged from less than $25,000 to over $200,000, with a mean of $76,000 – $100,000.	Parent-child interaction quality (88% of parents were mothers)	Observation during a 30m parent-child interaction (EAS)26	Baseline Reactivity (Lab-TAB27)	Salivary cortisol was collected after arrival at the lab in the morning (∼ 10:45 a.m.) or afternoon (∼1:15 p.m.), ∼25 m after arrival, ∼20 min after beginning of the Lab-TAB, and ∼ 20 min after the Lab-TAB.	Children who were extremely socially inhibited and who had higher-quality interactions with their parents exhibited smaller elevated in cortisol in response to the Lab-TAB than children who were extremely socially inhibited and had lower-quality interactions with their parents.	The quality of parent-child interaction was unrelated to HPA-axis reactivity among children who were not socially inhibited.
Kiel & Kalomiris, 2016	99	Toddlers (M=24.8 mos.) (SD=0.73 mos.) Income not reported. Socioeconomic status as rated on the Hollingshead scale28 ranged from 17 to 66, with a mean of 49.3 (SD=12.3), which corresponds to middle class.	Maternal comforting behavior	Observation during free play.	Baseline Reactivity (two fear-inducing tasks29)	Salivary cortisol was after the toddler acclimated to the lab (at least 20 m after arrival) and again 20 m after the completion of the second fear-inducing task.	Children of mothers who displayed higher levels of unsolicited or spontaneous comforting behaviors exhibited smaller elevations in cortisol in response to the fear-inducing tasks.	Findings for unsolicited comforting behaviors held after controlling for soliciting comforting behaviors and the level of fear toddlers displayed during the tasks.
Kryski et al., 2013	160	Preschoolers (M=3.6 yrs.) (SD= 0.2 yrs.) Income not reported. The mean socioeconomic status as rated on the Hollingshead scale was 46.1 (SD=10.3), which corresponds to middle class.	Parental positive and negative affect (96% of parents were mothers)	Observation (Teaching Tasks battery30)	Baseline Reactivity (Lab-TAB31 tasks)	Salivary cortisol was collected upon arrival at the lab, 30 min following the Stranger Approach task, 30 min after the Transparent Box task, and 20 min after the completion of the final Lab-TAB task.	Children rated lower on effortful control and whose parents displayed higher levels of positive affect exhibited lower levels of baseline cortisol. Children rated higher on effortful control and whose parents displayed higher levels of positive affect exhibited higher levels of baseline cortisol. Children rated lower on effortful control and whose parents displayed higher levels of negative affect exhibited higher levels of cortisol at baseline and in response to the Lab-TAB tasks.	Parental negative affect was unrelated to cortisol for children who were not related lower in levels of effortful control.
Laurent et al., 2016	100	Toddlers (1–3 years) Income not reported The majority (78%) of mothers had a college degree.	Maternal sensitivity and intrusiveness	Observation during free play and reunion following the Lab-TAB.	Baseline Reactivity (Lab-TAB)	Salivary cortisol was collected upon arrival at the lab, 15 m after arrival, and 20, 25, and 45 m after the Lab-TAB.	Children whose mothers displayed higher levels of sensitivity exhibited lower levels of cortisol at baseline and in response to the Lab-TAB, as well as more rapid recovery to baseline cortisol levels following the stressor. Children whose mothers displayed higher levels of intrusiveness exhibited slower recovery to baseline cortisol levels following the stressor.
Laurent et al., 2017	73	Infants (M=6 mos.) Median annual household income was between $10,000 and $19,999.	Maternal mindfulness	Self-report (IMP-I32)	Baseline Reactivity (SFP)	Salivary cortisol was collected upon arrival at the lab, immediately following the SFP, and again 15 and 30 m after the SFP.	Children of mothers who reported higher levels of mindful parenting behaviors exhibited smaller increases in cortisol in response to the SFP.
Luijk et al., 2010	369	Toddlers (M=14.7 mos.) (SD=0.9 mos.) Income not reported.	Attachment security and organization	Observation (SSP33)	Diurnal Reactivity (SSP)	To assess diurnal HPA-axis activity parents collected cortisol five salivary cortisol samples during a weekday at home.34 To assess HPA-axis reactivity, salivary cortisol samples were taken prior to the SSP, ∼10 m after the first separation episode, and ∼15 m later.	Children rated as disorganized exhibited a flattened diurnal pattern of cortisol relative to children who were not rated as disorganized. Children rated as insecure-resistant exhibited larger increases in cortisol in response to the SSP than children who were not rated as insecure-resistant.
Marceau et al., 2013	134	Children (M age at placement=7.11 days.) (SD=13.3 days.) Median annual household income was less than $15,000 for birth families and between $125,000 and $150,000 for adoptive families.	Overreactive parenting by adoptive parents	Self-report35	Variability in diurnal cortisol36	Salivary cortisol was collected ∼30 m after waking and before bed on three days when children were ∼4.5 yrs. of age.	Children who experienced high levels of prenatal risk (defined as a composite of substance use, depression, and anxiety) and whose adoptive mothers reported engaging in inconsistently-overreactive parenting across age exhibited lower levels of variability in diurnal cortisol. Children who experienced high levels of prenatal risk and whose adoptive fathers reported inconsistent overreactive parenting exhibited higher levels of variability in diurnal cortisol.	Among children who experienced low levels of prenatal risk there were no associations between inconsistent overreactive parenting and variability in diurnal cortisol.
Martinez-Torteya et al., 2014	153	Infants (M=7 mos.) Median annual household income was between $45,000 and $49,999, with over one-quarter of the sample (27%) below $20,000. Seventy-six percent of mothers had a history of childhood maltreatment.	Maternal positive behaviors, hostility, and overcontrolling behaviors37	Observation during teaching tasks and free play	Baseline Reactivity (SFP)	Salivary cortisol was collected prior to and 20, 40, and 60 m after the SFP.	Children whose mothers displayed higher levels of positive parenting behaviors were less likely to respond to the SFP (response was defined as an increase in cortisol of 10% or more over baseline). Children whose mothers displayed higher levels of hostility were more likely to respond to the SFP.
Martinez-Torteya et al., 201538	167	Infants & Toddlers (M=7–16 mos.) Median annual household income was between $45,000 and $49,999, with over one-quarter of the sample (27%) below $20,000. Seventy-six percent of mothers had a history of childhood maltreatment.	Maternal overcontrolling behaviors35	Observation during teaching tasks and free play	Baseline Reactivity (SFP)	At 7 mos., salivary cortisol was collected prior to and 20, 40, and 60 m after the SFP. At 16 mos., salivary cortisol was collected prior to and 20, 40, and 60 m after the SSP.	Maternal overcontrolling behaviors were associated with higher levels of baseline cortisol and reactivity.	Infants exhibited a cortisol response to the SSP at 16 mos., but not to the SFP at 7 mos.
Mills-Koonce et al., 201139	717	Infants (M=7.4 mos.) (SD=1.3 mos.) Mean income-to-needs ratio was 2.39 (SD=1.85) Toddlers (M=24.6 mos.) (SD=1.6 mos.) Mean income-to-needs ratio was 2.34 (SD=1.69)	Paternal sensitivity40 and negativity41	Observation (Infancy: free play) (Toddlerhood: puzzle task)	Baseline Reactivity (Infancy42) (Toddlerhood43)	Salivary cortisol was collected prior to challenge tasks, 20 m after the children’s peak emotional arousal, and 40 m after peak arousal.	Higher levels of paternal negativity were associated with higher levels of cortisol reactivity in infancy and higher overall levels of cortisol in toddlerhood.	Paternal sensitivity was not found to be associated with child cortisol levels. Paternal caregiving behaviors during infancy did not independently predict later cortisol activity during toddlerhood.
Morelius et al., 2007	22	Infants (M=10 wks.)44 Income not reported. For approximately half the sample (55%) the highest level of maternal education was elementary school. For the remainder (45%) it was high school.	Maternal sensitivity	Observation	Baseline Reactivity (Diaper change)	Salivary cortisol was collected prior to and 30 min after a diaper change.	Maternal sensitivity was not related to baseline cortisol levels or reactivity to diaper change.
Nachmias et al., 1996	77	Toddlers (M=18 mos.) Income not reported. Maternal education ranged from 12 to 21 years, with a mean of 15.8 years.	Attachment security and organization	Observation during the SSP	Baseline Reactivity	Salivary cortisol was collected upon arrival at the lab and ∼45 m after the beginning of the test session.	Elevations in cortisol were found only for inhibited toddlers in insecure mother-child attachment relationships.
Pendry & Adam, 2007	32	Kindergarteners (M=6.1 yrs.) Income not reported.	Maternal involvement and warmth	Self-report (IPPA45)	Diurnal	Salivary cortisol was collected at home immediately upon waking and just prior to bed on two consecutive days.	Children of mothers who displayed higher levels of involvement and warmth exhibited steeper declines in cortisol over the course of the day.	Maternal involvement and warmth were not associated with children’s waking cortisol levels or overall levels of cortisol.
Philbrook & Teti, 2016	82	Infants Time 1 (M=3.1 mos.) Time 2 (M=6.1mos.) Time 3 (M=9.2 mos.) Annual household income ranged from $9,500 to $300,000 with a median of $65,000.	Maternal emotional availability	Observation during bedtime (EAS46)	Nocturnal	Salivary cortisol was collected at 3, 6, and 9 mos. of child age just before the infant fell asleep and again upon waking the next morning.	Infants of mothers who displayed lower levels of emotional availability exhibited higher cortisol levels at bedtime at each age. Infants of mothers who responded more often to their non-distressed cues exhibited lower overall cortisol levels at each age. Infants of mothers who displayed higher levels of emotional availability exhibited low and stable nocturnal levels of cortisol.
Philbrook et al., 201447	167	Infants Time 1 (M=1 mo.) Time 2 (M=3 mos.) Annual household income ranged from $9,500 to $300,000 with a median of $65,000.	Maternal emotional availability	Observation (EAS)48 during free play and at bedtime	Nocturnal	Salivary cortisol was collected 3 times across the evening and following morning. The first sample was taken before dinner, the second just before the infant fell asleep, and the third sample when the infant awoke in the morning.	At T2, over half of the infants in the sample showed patterning49 in their cortisol across the night. At T2, infants of mothers exhibiting higher levels of emotional availability displayed lower cortisol at bedtime and lower overall levels of cortisol (as measured by the AUCG) and smaller increases in cortisol from baseline (AUCI).50	At T1, there were no significant differences in cortisol patterning across time with respect to maternal emotional availability at bedtime. At T2, higher levels of emotional availability were associated with a tendency for infants to show a decrease in cortisol across time at the trend level.
Rappolt-Schlichtman et al., 2009	60	Toddlers and Preschoolers (Range=2 – 4 yrs.) Most children (unspecified proportion) were from families living below or near the federal poverty level.	Child conflict with parents and teachers	Parent-child conflict: CPRS51 Teacher-child conflict: STRS	Diurnal	Salivary cortisol was collected four times in the morning on two days when children were in care: one day when they were in small groups, and another when they were engaged in normal classroom activities. Sample 1: 9:20 a.m.52 Range=8:30–10:15 a.m. Sample 2: 10:20 a.m. Range=9:20–11:25 a.m. Sample 3: 11:10 a.m. Range=9:45 a.m.-12:25 p.m. Sample 4: 12:15 p.m. Range=11:25 a.m.-1:00 p.m.	Cortisol levels decreased from the first to the third sample, and then increased from the third to the fourth sample. Children whose mothers reported higher levels of conflict exhibited more gradual declines in cortisol over the course of the day.
Roque et al., 2012	51	Toddlers (M=21.3 mos.) (SD=1.96 mos.) Income not reported. Maternal education ranged from 9 to 19 years, with a mean of 15.2 years (SD=3.04)	Attachment security and organization	Observation during mother-child interaction (AQS53)	Baseline Reactivity (Emotion Regulation Paradigm54)	Salivary cortisol was collected immediately before and 30 m after each episode of the emotion regulation paradigm.	Secure children showed significant increases in their cortisol levels after fear episodes and significant decreases after positive affect ones. Insecure children did not show significant differences in cortisol levels in any of the episodes.	No significant changes were found for frustration/anger episodes.
Schieche & Spangler, 2005	76	Infants & Toddlers Time 1 (M=12.4 mos.) Time 2 (M=22.4 mos.) Income not reported. Twelve percent of children were from lower class families, where class was defined as a composite of household income, paternal education, and paternal occupation.	Attachment security and organization; maternal supportive behaviors	Observation during SSP55 and problem-solving tasks56	Baseline Reactivity (problem-solving tasks)	Salivary cortisol was collected immediately before the problem-solving tasks and then 15 and 30 m after the task.	Among toddlers rated as behaviorally inhibited, response to the problem-solving task differed as a function of attachment classification: securely-attached toddlers exhibited a significant decrease in cortisol in response to the problem-solving task; comparable decreases were not observed among toddlers classified as insecure-ambivalent or disorganized.	Maternal supportive behaviors were not associated with children’s response to the problem-solving task.
Spangler et al., 1994	41	Infants (3, 6, 9 mos.) Income not reported. Over one-quarter (27%) of children were from families rated lower class based on paternal education, occupation, and income.	Maternal sensitivity	Observation during free play	Baseline Reactivity (Free play, diaper change57).	Salivary cortisol was collected before the play session, 15 m after free play session and 10 m after diaper change.	At 9 mos., infants of highly insensitive mothers had higher baseline cortisol levels. Cortisol increased significantly for infants with highly insensitive mothers at 3 mos. and 6 mos., but not at 9 mos.
Sturge-Apple et al., 2012a	201	Toddlers (M=27 mos.) Median annual household income was $18,300 and nearly all children were from families (99.5%) living at or below the federal poverty level.	Maternal emotional unavailability	Maternal report Observation during free play and a compliance task58	Baseline Reactivity (SSP, SPAT59)	Salivary cortisol was collected prior to the SSP and ∼25 and 50 m after the 5th episode. Cortisol was also collected prior to and ∼25 and 37 m after the SPAT.	Maternal emotional unavailability was negatively associated with cortisol change in response to the SSP. Maternal emotional unavailability predicted higher baseline cortisol levels on the day of the SSP. Maternal emotional unavailability was not significantly associated with baseline cortisol or cortisol reactivity the SPAT task.
Sturge-Apple et al., 2012b	201	Toddlers (M=27 mos.) Median annual household income was $18,300 and nearly all children were from families (99.5%) living at or below the federal poverty level.	Maternal harsh parenting (harsh, angry, critical, disapproving and/or rejecting behavior).	Maternal report Observation during free play and a compliance task60	Basal	Salivary cortisol was collected upon children’s arrival at the laboratory in the morning (∼9:30 a.m.)	Children of mothers who displayed higher levels of harsh parenting exhibited higher levels of baseline cortisol, but only if they were classified as belonging to the dove profile.61
Taylor et al., 2013	148	Toddlers, Preschoolers, & School-Aged Time 1 (T1) (M=18 mos.) Time 2 (T2) (M=30 mos.) Time 3 (T3) (M=72 mos.) Income not reported.	Maternal intrusive and over-controlling parenting	Observation (T1: free play) (T2: teaching paradigm task62, a clean-up task63)	Baseline Reactivity (Not-sharing task64)	Salivary cortisol was collected at T3, prior to the not-sharing task, 10 m after the task, and 20 m after the task	Intrusive-overcontrolling parenting at T2 predicted higher levels of children’s cortisol at T3.
Thomas et al., 2017	272	Infants (M=6 mos.) A small portion of the sample (9%) had annual household incomes of less than $40,000.	Mother-infant interaction quality	Observation while mothers taught their child a new task (PCITS65)	Baseline Reactivity (Lab-TAB66)	Salivary cortisol was collected prior to and 20 m following the Lab-TAB.	Higher mother-infant interaction quality was associated with smaller increases in cortisol in response to the Lab-TAB.	Interaction quality mediated the effects of perceived social support in pregnancy.
Thompson & Trevathan, 2008	63	Infants (M=13.2 wks.) (SD=0.84 wks.) Income not reported. Nearly one-third of the sample (31%) indicated that they face substantial financial stress.67	Maternal sensitivity	Observation during normal interaction68	Baseline Reactivity (maternal separation69)	Salivary cortisol was collected prior to and ∼20–30 m following initial separation.	Maternal sensitivity was not related to infants’ response to maternal separation.
Thompson et al., 2018	306	Preschoolers (M=37 mos.) (SD=0.84 mos.) Over one-quarter (29%) of children were from families with annual household incomes below or near the federal poverty level, while an additional 27% had incomes below the median.	Maternal warmth, negativity, limit setting, scaffolding, and responsiveness	Observation during mother-child interaction70	Diurnal	Salivary cortisol was collected twice a day (∼30 m after waking and ∼30 m before bed) on three consecutive days beginning when children were 36–40 mos. (T1) and at 9-mo. intervals thereafter (T2: 44–49 mos.; T3: 53–58 mos.; T4: 62–67 mos.)	Children of mothers who displayed stable (over time), higher levels of negativity exhibited lower levels of morning cortisol over time; children of mothers who displayed stable, higher levels of scaffolding exhibited higher levels of morning cortisol over time. Children of mothers who exhibited stable, higher levels of negativity and lower levels of responsiveness exhibited a flattened diurnal slope across ages.
Van Bakel & Riksen-Walraven, 2004	85	Toddlers (M=15.1 mos.) (SD=0.25 mos.) Income not reported.	Attachment security and organization.	AQS71	Baseline Reactivity (SSP)	Salivary cortisol was measured prior to and 3 m after the SSP.	More secure maternal-child attachment was associated with larger increases in cortisol in response to the SSP.
Zalewski et al., 2012a	78	Preschoolers Time 1 (T1) (M=36.6 mos.) Time 2 (T2) (M=42.0 mos.) Annual household income ranged from less than $20,000 to over $100,000, with 18% of the sample at or below the federal poverty level.	Maternal warmth, responsiveness, scaffolding, limit setting, and negative affect.	Observation at T1 during three tasks (restricted play, free play, Lego-building task)	Diurnal	Salivary cortisol was collected by at home in the morning and evening on two subsequent days. Morning: 30 m after waking Evening: 30 m before bedtime	Higher levels of maternal negativity were associated with a higher likelihood that children would exhibit a blunted pattern of diurnal cortisol. Children whose mothers displayed higher levels of warmth were less likely to demonstrate a blunted pattern of diurnal cortisol.	Maternal warmth and negativity mediated the associations between poverty and blunted cortisol at T2.
Zalewski et al., 2012b	306	Preschoolers (M=37.0 mos.) (SD=0.84 mos.) Nearly one-third of children came from households with annual household incomes at or below 150% of the federal poverty level.	Maternal warmth, responsiveness, scaffolding, limit setting, and negative affect.	Observation during three tasks (restricted play, free play, Lego-building task)	Diurnal	Salivary cortisol was collected at home in the morning and evening on three subsequent days. Morning: 30 m after waking Evening: 30 m before bedtime	Higher levels of maternal warmth were associated with higher levels of morning cortisol. Higher levels of maternal responsiveness were associated with steeper diurnal declines in cortisol, whereas higher levels of negativity were associated with flatter diurnal patterns.

¹Note that the sample employed in this study was the same as the sample employed by Albers et al. (2008).

²Video-Feedback Intervention to Promote Positive Parenting and Sensitive Discipline (VIPP-SD; Van Zeijl et al., 2006)

³Strange Situation Paradigm (Ainsworth, Blehar, Waters, & Wall, 1978).

⁴The sample for this study was drawn from the Family Life Project (FLP), as were the samples by Berry et al. (2014, 2016), Blair et al. (2008, 2015), Blair, Granger et al. (2011), Blair, Raver et al. (2011), and Mills-Koonce et al. (2015).

⁵Cox, Paley, Burchinal, & Payne, 1999

⁶Goldsmith & Rothbart, 1991

⁷Three tasks: a mask presentation challenge, a barrier challenge, and an arm restraint challenge.

⁸Two tasks: a toy removal challenge and a mask presentation challenge.

⁹Sensitivity, detachment, positive regard, animation, and stimulation.

¹⁰Intrusiveness and negative regard.

¹¹Some mothers used emotional withdrawal as a disciplinary tactic; others were withdrawn due to depression. Groups were combined for analysis.

¹²Conflict tactics scale (CTS; Straus, 1979)

¹³Strange Situation (Ainsworth, Blehar, Waters, & Wall, 1978).

¹⁴80% of children had experienced institutional care.

¹⁵Egeland et al., 1995.

¹⁶Laboratory Temperament Assessment Battery (Goldsmith, Reilly, Lemery, Longley, & Prescott, 1995), from which 12 tasks were selected.

¹⁷This task was developed by Kryski and colleagues (2011) and is based on a modified version of Lewis and Ramsay’s (2002) matching task.

¹⁸Conventional Still-Face Paradigm (SFP; Tronick, Als, Adamson, Wise, & Brazelton, 1978), or altered Still-Face Paragraph with touch (SFP+T).

¹⁹Mothers and children were asked to work together on a series of difficult puzzles and games (Saltaris & Samaha, 1998)

²⁰Foster children were divided into two groups, and of which was assigned to the intervention. Third group were control children residing with their families.

²¹Austin, 2003.

²²The Parenting Scale (Arnold, O’Leary, Wolff, & Acker, 1993) and Parenting Dimensions Inventory (Power, 1993) were used for Sample 1. Child Rearing Practices Report (Block, 1981) and Responses to Children’s Emotions (Hastings & De, 2008) were used for Sample 2. Parenting Styles & Dimensions Questionnaire (Robinson, Mandleco, Olsen, & Hart, 2001) were used for Sample 3.

²³Strange Situation Paradigm (Ainsworth, Blehar, Waters, & Wall, 1978).

²⁴A stranger approach task, a stranger working task, clown, puppet show, robot, and spider tasks.

²⁵Q-Sort (Waters & Deane, 1985).

²⁶Emotional Availability scales (EAS; Biringen, Robinson & Emde, 1998).

²⁷Laboratory Temperament Assessment Battery (Lab-TAB tasks; Gold-smith, Reilly, Lemery, Longley & Prescott, 1995).

²⁸Hollingshead’s Four Factor Index of Social Status (Hollingshead, 1975)

²⁹Including a Clown and a Spider episode (see Kiel & Buss, 2012).

³⁰Teaching Tasks battery (Egeland, Weinfield, Hiester, Lawrence, Pierce & Chippendale, 1995) included book reading, block building, naming objects with wheels, matching shapes, completing a maze using an etch-a-sketch, and gift presentation.

³¹Laboratory Temperament Assessment Battery (Lab-TAB tasks; Goldsmith, Reilly, Lemery, Longley & Prescott, 1995).

³²Interpersonal Mindfulness in Parenting – Infant Version (Duncan, 2007).

³³Strange Situation Paradigm (Ainsworth, Blehar, Waters, & Wall, 1978).

³⁴Samples were collected upon waking, 30 m after waking, between 11 am and 12 pm, between 3 and 4 pm, and at bedtime.

³⁵Using a subscale of the Parenting Scale (Arnold et al., 1993) at 9, 18, 27 mos. and 4.5 yrs. of child age.

³⁶Cortisol variability measures the extent to which an individual’s cortisol deviated from its usual diurnal pattern. It is operationalized as the residual cortisol circulating in the system after accounting for individual-specific diurnal declines.

³⁷Assessed using the MACY infant-parent coding system, (Earls, Beeghly, & Muzik, 2009).

³⁸This study employed a super-set of the sample reported on in Martinez-Torteya et al. (2014).

³⁹Used a sub-set of the FLP sample (fathers) also used by Berry et al. (2014, 2016, 2017) Blair et al. (2008, 2015), Blair, Granger et al. (2011), and Blair, Raver et al. (2011).

⁴⁰Sensitivity, detachment (reverse-coded), stimulation, positive regard, and animation.

⁴¹Intrusiveness and negative regard.

⁴²Three tasks: a mask presentation challenge, a barrier challenge, and an arm restraint challenge.

⁴³Two tasks: a toy removal challenge and a mask presentation challenge.

⁴⁴At age of first assessment.

⁴⁵Measures for parental involvement was designed for the purposes of this study, and warmth was measured by the Inventory of Parent and Peer Attachment (IPPA; Armsden & Greenberg, 1987)

⁴⁶Emotional availability scales (Biringen, Robinson, Emde, 1998).

⁴⁷Used a super-set of children employed in Philbrook & Teti (2016).

⁴⁸Emotional availability scales (Biringen, Robinson, Emde, 1998).

⁴⁹A decline in cortisol from late afternoon to bedtime, a gradual increase overnight, and sharp increase in the morning post-awakening were theorized to mark the presence of a nocturnal rhythm in infant cortisol.

⁵⁰AUCG: Area under curve with respect to ground; AUCI: area under curve with respect to increase.

⁵¹Child-Parent Relationship Scale (Pianta, 1992).

⁵²Times reported for each sample correspond to approximate time that sample was collected.

⁵³The Attachment Behavior Q-set (AQS, Waters, 1995).

⁵⁴Emotion Regulation Paradigm consists of three episodes: Fear, Positive Affect, and Frustration/Anger (Diener & Mangelsdorf, 1999).

⁵⁵Strange Situation Paradigm (Ainsworth, Blehar, Waters, & Wall, 1978).

⁵⁶The toddlers’ task-solving German versions of the 7-point scales, supportive presence and quality of assistance was developed by Matas et al. (1978).

⁵⁷At 6 and 9 mos., one saliva collection took place at participant’s home, the rest in the lab.

⁵⁸The Iowa family interaction rating scales (IFIRS; Melby & Conger, 2001)

⁵⁹Simulated Phone Argument Task (SPAT; Davies, Cummings, & Winter, 2004)

⁶⁰The Iowa family interaction rating scales (IFIRS; Melby & Conger, 2001)

⁶¹Profile classifications were made in accordance with Korte’s evolutionary model of temperament (Korte, Koolhaas, Wingfield, & McEwen, 2005), based on children’s temperament as rated by maternal report and coding of emotional response to four unfamiliar episodes. Children classified as doves were rated by their mothers as higher on falling reactivity and exhibited higher levels of inhibition, avoidance, and vulnerable affect in response to the unfamiliar episodes.

⁶²Adapted from Calkins & Johnson, 1998

⁶³Kochanska, Coy, & Murray 2001

⁶⁴Goldsmith, Reilly, Lemery, Longley, & Prescott, 1995. In this task, children played a card game with an experimenter that involved candy being unfairly divided based on what cards they received.

⁶⁵Parent Child Interaction Teaching Scale (PCITS; Oxford & Finlay, 2013).

⁶⁶Laboratory Temperament Assessment Battery (Lab-TAB tasks; Goldsmith & Rothbart, 1996).

⁶⁷13% indicated they “do not have enough money to meet basic needs” and 18% that they “barely pay the bills but manage on my own.”

⁶⁸A coding system developed by Isabella and Belsky (1991).

⁶⁹Maternal separation occurred during a learning and short-term memory exercise.

⁷⁰Adapted from the System for Coding Interactions and Family Functioning (SCIFF, Lindahl & Malik, 2000), the Parenting Style Ratings Manual (Cowan & Cowan, 1992), and the Parental Warmth and Control Scale-Revised (Rubin & Cheah, 2000)

⁷¹Attachment Q-Sort (Waters, 1995)

Table 2.

Studies of non-parental caregiver behaviors and HPA axis activity

Study	Participants		Assessment of Caregiving Behaviors		Measurement of HPA Axis Activity		Summary of Results
	N	Age & Socioeconomic Composition	Construct(s)	Measurement	Classification	Method & Timing	Major Findings	Notes
Albers et al., 2016	64	Infants (M=14.6 wks.) (SD=2.8 wks.) Income not reported. 34 mothers (53.1%) had a professional degree and 29 (45.3%) had at least one university degree.	Caregiver sensitivity and cooperation	Observation in the classroom	Diurnal	Salivary cortisol was obtained in the morning (10–10:30 a.m.) and afternoon (4–4:30 p.m.) while children were in care.	Quality of care was not associated with morning cortisol levels, afternoon cortisol levels, or diurnal slope.
Badanes et al., 2012	110	Preschoolers (M=4.03 yrs.) (SE=0.07 yrs.) Income not reported. Over half (56%) of children were from families that endorsed one or more sociodemographic risk factors.	Attachment to mother and primary teacher	AQS72	Diurnal	Salivary cortisol collected twice a day over three days. Morning: M = 9:50 a.m. SE=0.03 hrs. Afternoon: M = 3:34 p.m. SE=0.05 hrs.	More secure attachment to the lead teacher was associated with lower afternoon cortisol levels after controlling for morning levels.	Maternal attachment was not associated with cortisol levels.
Berry et al., 201473	1,235	Infants (∼7 mos.) Toddlers (∼15 & 24 mos.) Preschoolers (∼36 & 48 mos.) The majority (67%) of families had an annual household income of less than 200% of the federal poverty level.	Responsiveness	HOME74 (7, 15, 24, & 36 mos.)	Basal	Salivary cortisol collected once a day at ages 7 and 48 mos. 7 mos.: Mode=10:00 a.m. SD=2.88 hrs. 48 mos.: Mode=10:00 a.m. SD=2.91 hrs.	For children exposed to low levels of risk at home, more hours in care predicted higher cortisol levels at 48 months, after controlling for cortisol at 7 months. For children in higher-risk homes, more care was associated with lower levels of cortisol.	There was no effect of quality of care on these relations, nor was there a main effect for quality.
Berry et al., 201675	1,155	Infants (∼7 mos.) Toddlers (∼15 & 24 mos.) Preschoolers (∼36 & 48 mos.) The majority (67%) of families had an annual household income of less than 200% of the federal poverty level.	Responsiveness	HOME (7, 15, & 24 mos.)	Basal	Salivary cortisol collected once a day at ages 7, 15, & 24 months. Modal time of collection at each age was 10:00 a.m.	Less responsive caregiving behaviors were modestly associated with higher cortisol levels at 7 months, but not 15 or 24 months. For children from high-risk contexts, increases in child care exposure at 24 months were associated with decreases in children’s resting cortisol levels from 7 to 24 months.
Dettling et al., 2000a	40	21 preschoolers in home-based care (Mdn=51 mos.) (Range=40–69 mos.) 19 children not in care (Mdn=52 mos.) (Range=39–66 mos.) Median annual household income was $61,000–70,000.	Stimulation Attention	ORCE76	Diurnal	Salivary cortisol was collected twice on two days when children were or were not in care. Morning: 10:30–10:45 a.m. Afternoon: 3:30–3:45 p.m.	Children who experienced higher levels of stimulation and attention while in care were less likely to exhibit a rising pattern of cortisol over the course of the day.	Quality of care settings was within a truncated range (“good” to “excellent”).
Dettling et al., 2000b77	61	21 preschoolers in home-based care (Mdn=51 mos.) (Range=40–69 mos.) 21 preschoolers in center-based care (Mdn=52 mos.) (Range=35–63 mos.) 19 children not in care (Mdn=52 mos.) (Range=39–66 mos.) Median annual household income was $61,000–70,000.	Stimulation Attention	ORCE	Diurnal	Salivary cortisol was collected twice on two days when children were or were not in care. Morning: 10:30–10:45 a.m. Afternoon: 3:30–3:45 p.m.	In both caregiving settings, children who experienced higher levels of stimulation and attention while in care were less likely to exhibit a rising pattern of cortisol over the course of the day. In both caregiving settings, children who experienced higher levels of stimulation and attention exhibited steeper decreases in cortisol throughout the day.	Morning and afternoon cortisol levels were similar for children in home- and center-based care.
Groeneveld et al., 2010	102	45 preschoolers in center-based care (M=32.0 mos.) (SD=4.4 mos.) Income not reported. Average maternal education was 13.7 years. 71 preschoolers in home-based care (M=29.2 mos.) (SD=6.3 mos.) Income not reported. Average maternal education was 13.6 years.	Sensitivity	Seven-point scale78	Diurnal Overall (AUCg)79	Salivary cortisol was collected four times on one day when children were in care and one day when they were not. Sample 1: M=6:57 a.m. SD=28 min Sample 2: M=10:59 a.m. SD=5 min Sample 3: M=3:10 p.m. SD=22 min Sample 4: M=6:06 p.m. SD=35 min	Children in home-based care who experienced higher sensitivity exhibited lower overall levels of cortisol than children who experienced lower levels of sensitivity.	The relation between sensitivity and cortisol was not observed for children in center-based care.
Gunnar et al., 2010	151	Preschoolers (M=3.81 yrs.) (SD=0.23 yrs.) Mean annual household income was $51,000-$76,000.	Warmth/supportive Intrusive/controlling	M-ORCE80	Diurnal	Salivary cortisol was collected twice a day on two days when the child was in care and two days when the child was not in care. Morning: 10:00–11:00 a.m. Afternoon: 3:00–4:00 p.m.	Cortisol levels were higher among children rated as receiving higher levels of intrusive/controlling care.
Hatfield et al., 2013	63	Preschoolers (M=53.9 mos.) (SD=5.4 mos.) Nearly one-quarter (23%) of children came from families with annual household incomes of less than $24,000. Another 38% of children were from families with incomes between 24,000 and $47,999.	Emotional support	CLASS Pre-K81	Diurnal	Salivary cortisol was collected three times a day on two days when the child was in care. After arrival: M=8:29 a.m. Range=8:11–8:49 a.m. After morning play: M=10:48 Range=9:59 a.m.-12:10 p.m. Late afternoon: M=3:27 pm Range=1:53–4:41 p.m.	After controlling for structural quality, children in classrooms rated as featuring higher levels of emotional support demonstrated a decline in cortisol over the day.	Relations were not observed between classroom organization and cortisol or instructional support and cortisol.
Lisonbee et al., 2008	191	Preschoolers (M=53.0 mos.) (SD=4.1 mos.) Income not reported. Occupational prestige, based on parents’ reported employment, ranged from 25 (unskilled labor) to 93 (professional), with a median of 68.82	Child relationship with teacher (Clinginess) (Conflict) (Closeness)	Teacher-report: STRS83 Child-report: FAMST84	Diurnal Reactivity	Diurnal Salivary cortisol was collected twice on two days when the child was in care. Morning: M=9.00 a.m. Range=8:00–11:00 a.m. Afternoon: M=Not reported Range=1:00–4:00 p.m. Reactivity Salivary cortisol was collected three times over a 50 min procedure: prior to a set of challenging tasks, immediately following the tasks, and immediately following a 20-min interaction with their teacher	Higher levels of teacher-reported clinginess were associated with higher afternoon cortisol levels and larger increases in cortisol from morning to afternoon. Higher levels of teacher-reported conflict and lower levels of teacher-reported closeness were associated with larger increases in cortisol during the teacher-child interaction.	Results were robust to child gender, temperament, family socioeconomic status, and classroom characteristics (group size and teachers’ insensitivity at the classroom level85).
Rappolt-Schlichtman et al., 2009	60	Toddlers and Preschoolers in poverty (Range=2–4 yrs.) Most children were from families living below or near the federal poverty level.	Child conflict with parents and teachers	Parent-child conflict: CPRS86 Teacher-child conflict: STRS	Diurnal	Salivary cortisol was collected four times in the morning on two days when children were in care: one day when they were in small groups, and another when they were engaged in normal classroom activities. Sample 1: 9:20 a.m.87 Range=8:30–10:15 a.m. Sample 2: 10:20 a.m. Range=9:20–11:25 a.m. Sample 3: 11:10 a.m. Range=9:45 a.m.-12:25 p.m. Sample 4: 12:15 p.m. Range=11:25 a.m.-1:00 p.m.	Cortisol levels decreased from the first to the third sample, and then increased from the third to the fourth sample. When placed in small-group settings, children with higher levels of teacher conflict exhibited higher levels of cortisol. This was not observed when children were engaged in normal classroom activities
Sims et al., 2006	117	Preschoolers (Range=3–6 yrs.) Thirty-nine percent of children came from families with annual household incomes less than $41,000.	Warmth88	QIAS89	Diurnal	Salivary cortisol collected twice a day over three days.	Children attending high-quality centers exhibited a significant decline in cortisol across the day, while those attending satisfactory or unsatisfactory centers did not. Children attending unsatisfactory centers had higher mid-afternoon cortisol levels than their peers attending high-quality centers at a level approaching significance.
Watamura et al., 2009	65	Preschoolers (Range=3–6 yrs.) Three children (5%) came from families with annual household incomes of less than $25,000.	Sensitivity Classroom climate	M-ORCE	Diurnal	Salivary cortisol was collected three times a day on two days when the child was in care and two when they were not. Care Days: Morning (at care): Mdn=10:48 a.m. Range=9:16–11:23 a.m. Afternoon (at care): Mdn=3:48 p.m. Range=3:09–4:40 p.m. Evening (at home): Mdn=8:01 p.m. Range=7:00–10:35 p.m. Home Days: Morning: Mdn=10:07 a.m. Range=8:50–11:55 a.m. Afternoon: Mdn=3:44 p.m. Range=1:51–6:00 p.m. Evening: Mdn=8:06 p.m. Range=7:05–10:00 p.m.	Children exhibited an increase in cortisol on days when they were in care, relative to days when they were not in care. However, the magnitude of the increase and the proportion of children exhibiting the increase were smaller than those exhibited in previous studies where the quality of care was more variable. Children in classrooms rated as having more positive climates exhibited patterns of change in cortisol from morning to afternoon while in care that more similar to that observed when they were at home, as compared to children in classrooms rated as having more negative climates.	All centers were rated to be of excellent quality according to the ECERS-R.

⁷²Attachment Q-Sort (Waters, 1995)

⁷³Sample drawn from the FLP sample also used by Berry et al. (2016, 2017) Blair et al. (2008, 2015), Blair, Granger et al. (2011) Blair, Raver et al. (2011), and Mills-Koonce et al. (2011).

⁷⁴Home Observation for Measurement of the Environment scale (Caldwell & Bradley, 1984)

⁷⁵Sample drawn from the FLP sample also used by Berry et al. (2014, 2017) Blair et al. (2008, 2015), Blair, Granger et al. (2011) Blair, Raver et al. (2011), and Mills-Koonce et al. (2011).

⁷⁶Observational Rating of the Caregiving Environment, 56-month version (NICHD, 1996).

⁷⁷Used a super-set of the sample employed by Dettling et al. (2000a).

⁷⁸Scale was developed by the Dutch Consortium for Child Care Research (NCKO; De Kruif et al., 2007).

⁷⁹Area under the curve with respect to ground (Pruessner, Kirschbaum, Meinlschmid, & Hellhammer, 2003).

⁸⁰Modified Observational Rating of the Caregiving Environment (Gunnar et al., 2001).

⁸¹Pianta, La Paro, & Hamre, 2008.

⁸²Occupational prestige rated using a scale developed by Entwisle & Astone, 1994.

⁸³Student-Teacher Relationship Scale (Pianta et al., 1995).

⁸⁴Feelings About My School and Teacher Interview (Payne, 2003).

⁸⁵As measured using the Caregiver Interaction Scale (CIS; Arnett, 1989).

⁸⁶Child-Parent Relationship Scale (Pianta, 1992).

⁸⁷Times reported for each sample correspond to approximate time that sample was collected.

⁸⁸Warmth was operationalized as a combination of Positive guidance communication, cultural competence, and equitable treatment. However, the QIAS also includes nine additional subscales that index aspects of structural quality.

⁸⁹Quality Improvement and Assurance System (National Childcare Accreditation Council, 2001).

Despite these efforts to facilitate the comparison of results, the studies included in the review were quite heterogeneous in terms of sample size and composition, dimensions of caregiver behaviors assessed and methods of assessment, and the aspects of HPA-axis activity that were measured and how. This heterogeneity precluded the possibility of conducting a meta-analysis. Therefore, we instead provide a systematic narrative synthesis of the literature that emphasizes points of convergence within and between the parental and non-parental caregiving literatures.

Results

Parental Caregiving Behaviors and HPA-axis Activity in Early Childhood

Decades of research from animal models indicate that the early caregiving environment plays a critical role in the development and canalization HPA-axis activity (Champagne et al., 2008; Liu, Diorio, Day, Francis, & Meaney, 2000), and over the past 20 years a complementary body of research has emerged that testifies to the capacity of parenting behaviors to influence HPA-axis activity in young children. Applying the search criteria detailed above yielded 67 studies for inclusion in the review, 55 of which concerned parental caregiving behaviors (see Figure 1 for a summary of the search process and Table 1 for a list of studies featuring parental caregiving behaviors included in the review).

Figure 1.

Summary of the results of the literature search. Note that: 1) “report characteristic restrictions” (right-most column, top-most box) refer to inclusion criteria based on report characteristics (e.g., published in English); the text following the heading (e.g., “no caregiver behaviors assessed”) in the box in the right-most column, third from the top lists the reasons studies were excluded following full-text review; and 3) that two studies featured parental and non-parental caregiving and are therefore included in Tables 1 and 2.

Four studies documented an association between maternal sensitivity and reactivity of the HPA axis in response to challenge, and three of these found that lower levels of sensitivity were associated with larger increases in cortisol in response to challenge. For example, Spangler and colleagues (1994) found that infants of mothers who displayed very low levels of sensitivity exhibited higher cortisol levels in response to a diaper change and free-play session with their mothers at 3 and 6 months than infants of more sensitive mothers (Spangler, Schieche, Ilg, Maier, & Ackermann, 1994). Studies with slightly older infants (9 months; Grant et al., 2009) and toddlers (1 – 3 years; Laurent, Harold, Leve, Shelton, & Van Goozen, 2016) reported similar results in response to the still-face procedure (Tronick, Als, Adamson, Wise, & Brazelton, 1978) and the Laboratory Temperament Assessment Battery (Lab-TAB; Goldsmith, Reilly, Lemery, Longley, & Prescott, 1995). Only in one study were lower levels of sensitivity associated with smaller increases in cortisol in response the challenge (the Strange Situations Procedure; Ainswoth, Blehar, Waters, & Wall, 1978), and in this case all children had been adopted internationally (DePasquale, Raby, Hoye, & Dozier, 2018).

Researchers have also examined the relation between maternal sensitivity and other aspects of HPA-axis activity, with six studies reporting an association between sensitivity and levels of cortisol under homeostatic conditions (Note that here and throughout “basal” refers to cortisol collected outside of the context of a challenging task, while the term “baseline” is reserved to refer to cortisol levels obtained in the context of such a task, prior to the onset of the challenge. When a specific aspect of HPA-axis activity was assessed, such as waking or diurnal function, those qualifiers are used). Two studies (Spangler et al., 1994; Grant et al., 2009) found that at 9 months of age, infants of mothers who displayed lower levels of sensitivity exhibited higher levels of cortisol at baseline, and another (Laurent et al., 2016) found similar results among toddlers (ages 1 to 3 years). Kaplan and colleagues (2008) reported similar results among infants (~4 months), though in this study the relation between lower maternal sensitivity and higher baseline cortisol levels was contingent upon the mother having a psychiatric diagnosis (Kaplan, Evans, & Monk, 2008). Blair and colleagues expanded upon these results in a pair of longitudinal studies using data from the Family Life Project in which maternal sensitivity and baseline cortisol levels were assessed at various ages between at 7 and 48 months (Berry et al., 2017; Blair et al., 2011). Children of mothers who displayed lower levels of sensitivity at 7, 15, 24, 36, or 48 months of age exhibited higher levels of cortisol at that same age (Blair et al., 2011), and children of mothers who displayed lower levels of sensitivity across ages 7, 15, and 24 months of age exhibited higher, cross-age cortisol levels (Berry et al., 2017).

Differences in maternal sensitivity have also been linked to how quickly children’s cortisol levels return to baseline following a challenge. Albers and colleagues found that, following the mild challenge presented by a bath, infants and toddlers (~12 months) of mothers who exhibited higher levels of sensitivity displayed more rapid returns to baseline levels of cortisol observed prior to the bath than infants of mothers who exhibited lower levels of sensitivity (Albers, Riksen-Walraven, Sweep, & Weerth, 2008). Similar results were found among toddlers (1 – 3 years) following the Lab-TAB (Laurent et al., 2016).

Like some studies of non-parental caregiving (see below), in nine studies attachment bonds have been used as a proxy measure of caregiving behaviors in examinations of infants’ HPA-axis activity. A large body of research demonstrates that sensitive, responsive, and warm caregiving behaviors foster secure bonds between mothers and children (de Wolff & van IJzendoorn, 1997; Goldsmith & Alansky, 1987), and therefore mother-child attachment security and organization is often used as a proxy measure of maternal sensitivity. Two studies have found that younger (~15 months; Bernard & Dozier, 2010) and older toddlers (~19 months; Hertsgaard, Gunnar, Erickson, & Nachmias, 1995) rated as having a disorganized attachment bond with their mothers exhibit larger elevations in cortisol in response to the Strange Situations Procedure (SSP; Ainsworth, Blehar, Waters, & Wall, 1978). Four studies have found associations between insecurity and elevated cortisol responses to the SSP among infants (2 and 6 months – note that attachment classifications were made at 18 months; Gunnar, Brodersen, Nachmias, Buss, & Rigatuso, 1996), toddlers (~15 months, Luijk et al., 2010; ~18 months, Nachmias, Gunnar, Mangelsdorf, Parritz, & Buss, 1996), as well as parallel links between insecurity and higher levels of cortisol in response to a problem-solving task in toddlerhood (~22 months; Schieche & Spangler, 2005). It is, however, important to note that in two cases associations between insecurity and increased reactivity were contingent upon children being classified as fearful (Gunnar et al., 1996) or inhibited (Nachmias et al., 1996), and that in two other studies securely attached toddlers exhibited larger increases in cortisol in response to the SSP (Van Bakel & Riksen-Walraven, 2004) or to fear episodes from an emotion regulation task (Roque, Veríssimo, Oliveira, & Oliveira, 2012). The security and organization of mother-child attachment bonds have been linked to aspects of HPA-axis activity other than reactivity in two studies. Luijk and colleagues (2010) found that toddlers (~15 months) rated as disorganized exhibited flattened diurnal patterns of cortisol, while Johnson et al. (2018) found that toddlers (~15 months) who were insecurely attached displayed higher baseline levels of cortisol, but only if they were also from households at or near the federal poverty level.

Various aspects of parenting closely related to sensitivity – engagement, involvement, warmth and responsiveness – have also been associated with individual differences in HPA-axis activity in six studies. Using data drawn from the Family Life Project, Blair and colleagues (2008) found that higher levels of maternal engagement in infancy (~7 months) were associated with larger increases in cortisol in response to challenge in infancy and lower overall, baseline levels of cortisol (the aggregate of three cortisol samples taken) in toddlerhood (~15 months). Among older toddlers (~25 months), higher levels of unsolicited comforting behaviors by mothers were associated with smaller reaction to challenge (Kiel & Kalomiris, 2016). Where diurnal cortisol is concerned, Pendry & Adam (2007) found that higher levels of involvement were associated with lower diurnal levels of cortisol among kindergarteners (~6 years). In two studies of preschoolers, Zalewski and colleagues found that children whose mothers displayed higher levels of warmth at ~36 months were less likely to exhibit flattened patterns of diurnal cortisol at ~42 months (Zalewski, 2012a), and that higher levels of warmth were also associated with higher morning levels of cortisol (Zalewski, Lengua, Kiff, & Fisher, 2012b), which may permit more pronounced reductions in cortisol throughout the day (Miller et al., 2007). Indeed, Zalewski et al. (2012b) found that higher levels of maternal responsiveness were associated with these steeper diurnal declines in preschool, a pattern that Thompson and colleagues (2018) extended throughout preschool and into school age by following children in this same sample longitudinally (Thompson, Zalewski, Kiff, & Lengua, 2018).

Five studies have found associations between parents’ emotional availability and various aspects of children’s HPA-axis activity. Toddlers of mothers who reported engaging in frequent emotional withdrawal (either as a disciplinary tactic or due to depression) exhibited higher baseline levels of cortisol (Bugental et al., 2003), while toddlers of mothers who displayed lower levels of emotional availability exhibited smaller increases in cortisol in response to the SSP, but not to a simulated phone argument task (SPAT; Davies, Cummings, & Winter, 2004; Sturge-Apple, Davies, Cicchetti, & Manning, 2012a). Preschoolers whose mothers displayed lower levels of emotional availability exhibited lower waking cortisol levels, which, as noted above, may enhance the likelihood of flatter diurnal rhythms (Fisher, Serbin, et al., 2007). Philbrook and colleagues examined the association between maternal emotional availability and nocturnal levels of cortisol in infants (1 – 9 months) by sampling cortisol before bedtime and again when infants woke the next morning. At 1 month there was no association between emotional availability and bedtime or nocturnal levels of cortisol, but by 3 months infants of mothers who displayed lower overall levels of emotional availability exhibited higher levels of cortisol at bedtime, higher levels of cortisol over the course of the night, and larger increases in cortisol relative to the first sample (Philbrook et al., 2014). In the same sample the association between lower emotional availability and higher cortisol at bedtimes was also observed at 6 and 9 months, but there was also a longitudinal association between availability and cortisol, such that infants whose mothers displayed higher emotional availability over time exhibited low and stable nocturnal cortisol levels over time (Philbrook & Teti, 2016).

Finally, six studies have examined the association between maternal caregiving behaviors, that, while positive, either resist grouping with the constructs above or are composites of multiple positive behaviors. Where specific behaviors are concerned, mothers of infants (~6 months) who reported higher levels of mindfulness exhibited smaller increases in cortisol in response to the Still-Face Paradigm (SFP; Laurent, Duncan, Lightcap, & Khan, 2017), while children whose mothers exhibited stable, higher levels of scaffolding from preschool through the transition to school exhibited higher levels of waking cortisol across this period (Thompson et al., 2018). Higher levels of positive maternal behaviors, broadly defined, have been linked to lower baseline levels of cortisol contemporaneously (7 – 36 months; Blair et al., 2011a, using data drawn from the Family Life Project) and prospectively (7 – 48 months; Blair et al., 2011b, using data from the same project), as well as a reduced magnitude (Thomas, Letourneau, Bryce, Campbell, & Giesbrecht, 2017) and likelihood (Martinez-Torteya et al., 2014) of infant (6 – 7 months) reactivity to the Lab-TAB and SFP, respectively.

Negative Parental Caregiving Behaviors

In summary, the literature reviewed above clearly demonstrates that individual differences in various aspects of HPA-axis activity are associated with parenting behaviors that may be broadly construed as positive. However, parallel associations have also been observed between HPA-axis activity and a variety of negative parenting behaviors in ten studies. Using a sample reported on in an earlier study (Martinez-Torteya et al., 2014), Martinez-Torteya and colleagues (2015) found that infants whose mothers displayed higher levels of intrusive behaviors exhibited higher levels of baseline cortisol and larger elevations in cortisol in response to the SFP at ~7 months and the SSP at ~16 months, while Laurent and colleagues (2016) reported that toddlers (1 – 3 years) whose mothers were more intrusive took longer to recover to baseline cortisol levels following the Lab-TAB (Laurent et al., 2016). Higher levels of maternal intrusiveness in preschool (~30 months) also predicted higher levels of cortisol prior to and following a not-sharing task at ~72 months (Taylor et al., 2013). Interestingly, even when these maternal behaviors fell short of intrusive and were merely protective, higher levels of them were associated with larger increases in cortisol in response to novel tasks among toddlers (~24 months; Hutt, Buss, & Kiel, 2013).

Among preschoolers, higher levels of negative affect directed by the mother towards the child at 36 months were associated with higher diurnal levels of cortisol both concurrently (Zalewski et al., 2012b) and longitudinally at ~42 months (Zalewski et al., 2012a), an association that Thompson and colleagues observed through the transition to school (Thompson et al., 2018) using the same sample as Zalewski et al. (2012b). A similar construct, maternal hostility, was also linked to a higher likelihood of an increase in cortisol in response to the SFP among infants (~7 months; Martinez-Torteya et al., 2014) and a larger increase in cortisol in response to the Lab-TAB among preschoolers whose mothers had been depressed in the first year of life (Dougherty, Tolep, Smith, & Rose, 2013). Toddlers who were classified as “doves” based on a temperament characterized by inhibition and avoidance (Korte, Koolhaus, Wingfield, & McEwen, 2005) and whose mothers engaged in higher levels of harsh parenting exhibited higher basal levels of cortisol (Sturge-Apple et al., 2012b, using the same sample employed in Sturge-Apple et al., 2012a), while preschoolers of mothers who reported in engaging in higher levels of punishment exhibited larger increases in cortisol in response to interactions with unfamiliar adults (Hastings et al., 2011). Finally, Rappolt-Schlichtman and colleagues (2009) found that toddlers and preschoolers of mothers who reported higher levels of parent-child conflict exhibited more gradual declines in diurnal cortisol over the course of the day.

Paternal Caregiving Behaviors

Although this section of the paper ostensibly addresses “parental” caregiving behaviors and HPA-axis activity, thus far all the studies presented are concerned with maternal behaviors. This reflects the fact that, to date, only five studies have examined the association paternal behaviors and children’s HPA-axis activity, either independently or in conjunction with maternal parenting behaviors. For example, Dougherty and colleagues reported that preschoolers of parents exhibiting higher levels of hostility displayed higher levels of cortisol prior to and during a series of tasks drawn from the Lab-TAB, but their sample was comprised almost exclusively (96%) of mothers (Dougherty, Klein, Rose, & Laptook, 2011). Another study using Family Life Project data reported that toddlers (~24 months) who mount a strong emotional response to a fear-inducing mask and toy removal exhibit larger cortisol increases in response to the task if their parents are more sensitive (Blair et al., 2015), whereas preschoolers (~48 months) who are extremely socially inhibited exhibit smaller increases in cortisol in response to the Lab-TAB if they have more positive interactions with their parents (Kertes et al., 2009), but the parents in question were approximately 90% mothers in both cases. Preschoolers response to the Lab-TAB also differed as a function of parental affect and child effortful control, such that children who were rated as lower on effortful control and whose parents displayed higher levels of negative affect exhibited larger increases in cortisol in response to the Lab-TAB, but here too the sample was almost exclusively (96%) mothers (Kryski et al., 2013). Indeed, to date only one study has focused on paternal behavior and young children’s HPA-axis activity. Using a sub-sample drawn from the Family Life Project, Mills-Koonce and colleagues examined cortisol levels among infants and toddlers prior to and following challenge tasks that did not involve the fathers (Mills-Koonce et al., 2011). While paternal sensitivity was not related to cortisol levels at either age, higher levels on a composite measure of negative regard and intrusiveness in infancy were related to higher levels of cortisol in response to challenge in infancy, and higher overall levels of cortisol in toddlerhood.

Evidence from Experimental Studies

All of the studies reviewed above are correlational in nature, and while many accounted for alternative sources of individual differences in children’s HPA-axis activity, none can establish whether parenting behaviors caused those differences. Experimental studies employing randomized-control designs are better suited to address this question, and six such studies are reviewed here. Five of these involved assigning families at random to extended interventions designed to change parental caregiving behaviors over time, whereas one study employed a manipulation in the laboratory to examine the association between these behaviors and HPA-axis activity. Feldman and colleagues (2010) randomly assigned mother-infant dyads to participate in the still-face procedure as designed, or in a modified procedure in which mothers were required to maintain a still-face but were allowed to touch their child (Feldman, Singer, & Zagoory, 2010). Children assigned to the conventional still-face procedure exhibited larger elevations in cortisol than their peers, and to the extent the conventional still-face more closely emulates lower maternal sensitivity than the altered procedure, these results are broadly consistent with those reported by Grant et al. (2009) and Spangler et al. (1994).

Fisher and colleagues worked with children who had a history of maltreatment and their foster families, examining HPA-axis activity among children of foster parents randomly assigned to an intervention designed to improve the quality of caregiving (Fisher, Stoolmiller, Gunnar, & Burraston, 2007). Over the course of a year, children of foster parents receiving the intervention came to exhibit diurnal patterns of cortisol similar to those exhibited by comparison children who had not been maltreated. In contrast, children of foster parents who did not receive the intervention exhibited an increasingly ‘flattened’ pattern of diurnal cortisol activity characteristic of many children reared in foster care, in which cortisol levels do not demonstrate their typical decline over the course of the day. A subsequent study conducted by Fisher and his colleagues employed this same intervention, but examined its effects on children’s diurnal cortisol following a change in foster-care placement (Fisher, Van Ryzin, & Gunnar, 2011). Children placed with families that had not received the intervention exhibited increasing levels of diurnal cortisol in the weeks following the new placement; children placed with families that had received the intervention did not. Bakermans-Kranenburg and colleagues reported similar results in a study in which mothers of toddlers were randomly assigned to receive a program designed to enhance maternal sensitivity and improve discipline practices (Bakermans-Kranenburg, van IJzendoorn, Mesman, Alink, & Juffer, 2008). Program participation caused reductions in children’s diurnal cortisol, though this effect was conditional upon children possessing a particular allele (DRD4 7-repeat).

Bugental and colleagues randomly assigned high-risk families (many of whom were in poverty) to a year-long intervention designed to increase caregivers’ parenting knowledge and skills during the child’s first year of life, and assessed children’s basal cortisol levels annually over three years (Bugental, Schwartz, & Lynch, 2010). Children whose families received the intervention exhibited lower cortisol levels one year and again three years after participation, and, in contrast to their control-group peers, these children exhibited declines in cortisol over time. A similar study was conducted by Cicchetti and colleagues, though in this case the families of infants and toddlers who had been maltreated were assigned to an intervention designed to promote sensitive parental caregiving behaviors (Cicchetti, Rogosch, Toth, & Sturge-Apple, 2011). After six months, maltreated children who were receiving the intervention exhibited basal cortisol levels that were no different than those of their non-maltreated peers. In contrast, maltreated children who were not receiving the intervention exhibited progressively lower cortisol values over time that were significantly different than those observed among maltreated children receiving the intervention and their non-maltreated peers. Note that other intervention studies designed to promote positive parenting behaviors and/or reduce negative behaviors have also emphasized child-level outcomes (e.g., teaching children self-regulatory skills), and that these programs have also been found to alter children’s HPA-axis activity (see Dozier, Peloso, Lewis, Laurenceau, & Levine, 2008); for a broader review of the intervention literature that includes studies outside the inclusion and exclusion criteria used here, see Slopen, McLaughlin, and Shonkoff (2014).

Summary

There is ample evidence that certain parenting behaviors, such as reduced levels of sensitivity (Albers et al., 2008; Berry et al., 2017; Blair et al., 2011; Grant et al., 2009; Laurent et al., 2016; Spangler et al., 1994) and increased hostility (Dougherty et al., 2013; Martinez-Torteya et al., 2014), are associated with individual differences in various aspects of HPA-axis activity. Moreover, the literature indicates that poverty may increase the likelihood that parents engage in these behaviors (Holochwost et al., 2016; Popp, Spinrad, & Smith, 2008). Together, these findings outline a mediation model in which parenting behaviors mediate the relation between poverty and HPA-axis activity, though this model that has been tested only in a small number of studies to date. Blair and colleagues found that maternal caregiving mediated associations between poverty, as indexed by families’ income-to-needs ratio, and children’s baseline cortisol levels at 7-, 15-, and 24-months of age. Specifically, higher income-to-needs ratios were associated with higher levels of maternal sensitivity, which were in turn associated with lower levels of baseline cortisol levels at 7, 15, and 24 months (Blair et al., 2011). Zalewski and colleagues reported similar results among preschoolers: higher levels of poverty at 36 months were associated with lower levels of warmth and higher levels of negativity at 36 months; these, in turn, predicted higher diurnal levels of cortisol at 42 months (Zalewski et al., 2012a).

Non-Parental Caregiving Behaviors and HPA-axis Activity in Early Childhood

Most studies of non-parental caregiving behaviors and the HPA axis have focused on diurnal activity. As Berry and colleagues (2016) noted, children attending care are ideal candidates for participation in studies employing elegant, within-person designs in which diurnal levels of cortisol on days when children are in non-parental care settings can be compared with levels on days when children are at home. In general, these studies have found that when children are in care they exhibit higher levels of cortisol over the course of the day (that is, “flatter” or less steeply negative diurnal trajectories of cortisol) than when they are at home (Dettling, Gunnar, & Donzella, 2000a; Tout et al., 1998; Watamura et al., 2002), though, as we note below, this pattern may not hold for children in poverty (Rappolt-Schlichtman et al., 2009). Various child factors are associated with diurnal levels of cortisol while in care. For example, younger children (ages 3 to 4) exhibit higher diurnal cortisol than older children (ages 7 to 8; Dettling et al., 2000a; Watamura et al., 2002), as do children with poorer emotional or behavioral regulation (Dettling et al., 2000a; Dettling, Parker, Lane, Sebanc, & Gunnar, 2000b; Tout et al., 1998) and children who exhibit higher levels of social fear or anxiety (Watamura et al., 2002).

However, diurnal activity of the HPA axis is also linked to the quality of the caregiving environment – indeed, one meta-analysis found that the size of the effect for quality of care on HPA-axis activity was much larger than the effect for age (Geoffroy et al., 2006). Structural quality includes the ratio of adults to children, staff qualifications, and physical aspects of the classroom environment (Zaslow, Tout, & Martinez-Beck, 2010), and even when overall structural quality is high (within the “good” to “excellent” range on the ECERS), children in preschool exhibit increases in cortisol levels throughout the school day (Tout et al., 1998). This finding has been replicated in multiple subsequent studies of classrooms that were also rated high on structural quality (Watamura et al., 2002, 2003, 2009), and expanded to include centers of lower structural quality. For example, in an international sample, Groeneveld and colleagues (2010) found that children in center-based care rated lower in structural quality (on the ECERS-R; Harms, Clifford, & Cryer, 1998) also exhibited increases in cortisol over the course of the day. Sajaniemi and colleagues (2011) reported similar results using a different sampling protocol in which cortisol was collected on days when children were at home (when the first, second, and final samples were taken) and in school (when the third and fourth samples were taken). They observed higher levels of diurnal cortisol among children in classrooms with lower ratings on various aspects of structural quality, and also found that children in classrooms with higher quality ratings exhibited steeper declines in cortisol over the course of the school day (Sajaniemi et al., 2011). Legendre and colleagues (2003) found that specific aspects of structural quality were differentially associated with diurnal cortisol. Smaller peer groups and more space for free play were associated with lower diurnal levels, while a larger number of caregivers was associated with higher diurnal levels, a finding that the authors attributed to the potential for a large number of caregivers to interfere with the child’s ability to form strong relationships with any particular caregiver.

Process quality – which includes non-parental caregiving behaviors – has also linked to HPA-axis activity while children are in care. Though effective classroom organization and management are hallmarks of high process quality, its “core feature” (Groenevald et al., 2010, p. 503) is a pattern of caregiver-child interactions that foster a climate of emotional support, including sensitive and warm caregiving behaviors (Burchinal et al., 2000; Hamre & Pianta, 2007). Our search of the literature identified 13 studies that have examined the association between non-parental caregiving behaviors and HPA-axis activity (see Figure 1 and Table 2). Eleven of these studies examined the relation between caregiver behaviors and diurnal levels of cortisol, and have done so in two ways: indirectly, by using questionnaires or attachment classification to assess relationships between teachers and children, and directly, through observational measurement of teacher-child interactions.

A number of studies have suggested that young children form attachment bonds with their non-parental caregivers (Howes & Hamilton, 1992; Pianta, Nimetz, & Bennett, 1997; Howes, Galinsky, & Kontos, 1998), and therefore the attachment bond between a non-paternal caregiver and their student may serve as a valid proxy measure of caregiving behavior (Badanes et al., 2012, citing Dettling et al., 2000a). Badanes and colleagues (2012) found that preschoolers rated as having a more secure attachment with their lead teacher exhibited lower levels of diurnal cortisol (i.e., lower afternoon levels after controlling for morning levels), suggesting that a secure attachment relationship helps children cope with potentially stressful situations in the classroom (Stansbury & Gunnar, 1995). Lisonbee and colleagues (2008) reported similar results when examining the association between preschoolers’ cortisol and teachers’ perceptions of the quality of their relationship with each student (as measured by the Student-Teacher Relationship Scale, or STRS; Pianta et al., 1995). Students who were rated by teachers as more “clingy” exhibited higher afternoon levels of cortisol and larger increases in cortisol over the course of the day. This association was robust to children’s gender, temperament, and, perhaps most important, teachers’ observed sensitivity (or lack thereof) when interacting with the students in their class as a whole.

Studies of the association between diurnal cortisol and direct observational measures of teacher behavior have employed various indices of caregiving behaviors, ranging from very broad to quite specific. Perhaps the broadest measure of caregiving behaviors was the Quality Improvement and Assurance System (QIAS; National Childcare Accreditation Council, 2001) employed by Sims and colleagues (2006), which included caregivers’ warmth, positive guidance, communication, cultural competence, and equitable treatment as components in a global index of quality that also assessed structural aspects of care. Australian preschoolers attending government-funded child care centers rated as high quality on this measure exhibited the typical diurnal decrease in cortisol, while those attending centers rated as satisfactory or unsatisfactory did not. However, given that the index conflated process and structural quality, it is impossible to determine to what extent variations in diurnal cortisol were attributable to caregiving behavior.

Fortunately, a number of studies have examined the association between non-parental caregiving behavior and diurnal cortisol while accounting for structural aspects of quality. For example, Dettling and colleagues (2000a) measured diurnal cortisol among preschoolers on days when they were and were not in home-based care, but included in their sample only those settings where the structural quality was rated as “good” or “excellent” (on the ECERS-R). Students of caregivers who displayed higher levels of stimulation and attention (as measured by the Observational Rating of the Caregiving Environment, or ORCE; NICHD, 1996) were less likely to exhibit an increasing pattern of diurnal cortisol. A follow-up study extended this finding to children attending center-based care (by adding a sub-sample of children from that setting to the sample featured in Dettling et al. 2000a), but also reported that children who experienced higher levels of stimulation and attention exhibited more pronounced declines in cortisol over the course of the day (Dettling et al., 2000b). Using a modified version of the same measure employed by Dettling et al. (2000a & b), the M-ORCE (Gunnar, Kryser, Phillips, & Vandell, 2001), Watamura and her colleagues found that whereas preschoolers exhibited an increase in cortisol on days when they were in care regardless of quality, relative to days when they were not in care, children in classrooms rated as having more positive climates exhibited diurnal patterns of cortisol that were more similar to the patterns they exhibited on days when they were at home (Watamura et al., 2009). As in the studies conducted by Dettling et al., (2000a), structural quality was accounted for by including only children attending centers of a specified quality, though here centers had to be rated as “excellent” on the ECERS-R. In contrast, Hatfield and colleagues (2013) employed statistical controls to account for elements of structural quality in their analyses, allowing them to isolate the association between emotional support provided by caregivers and preschoolers’ diurnal cortisol. Children in classrooms rated as evidencing higher levels of emotional support exhibited a decline in cortisol levels over the course of the day. Moreover, even when structural quality of the classroom was high, if emotional support was low children exhibited increases in diurnal cortisol.

Other studies have examined the association between non-parental caregiving behaviors and diurnal cortisol without reference to structural quality. Groenevald and colleagues (2009) examined diurnal cortisol (defined as overall cortisol across four samples taken throughout the day) in preschoolers who were in home- or centered-based care. Among children in home-based care, higher levels of teacher sensitivity (as rated on a scale developed by the Dutch Consortium for Child Care Research; De Kruif et al., 2007) were associated with lower overall levels of cortisol; however, this association was not observed among children in center-based care. In another study of children in center-based care, no relation was found between teacher warmth (as rated by the M-ORCE; Gunnar et al., 2001) and cortisol. Higher levels of intrusive or controlling caregiving behaviors were associated with elevated diurnal cortisol, though these behaviors accounted for a modest portion (4%) of the variance in cortisol levels (Gunnar et al., 2010).

A smaller number of studies have examined the association between non-parenting caregiving behaviors and aspects of HPA-axis activity beyond diurnal function. Only one of these deliberately examined the association between these behaviors and children’s responses to challenge. Lisonbee and colleagues (2008) measured preschoolers’ cortisol three times over the course of a 50-min procedure: prior to engaging in a set of challenging tasks (time 1), immediately following the tasks (time 2), and immediately following a 20-min interaction with their teacher (time 3). Students who were rated by teachers as displaying higher levels of teacher-child conflict and lower levels of teacher-child closeness exhibited larger increases in cortisol from time 2 to time 3. As with the findings for diurnal cortisol, these associations were robust to child gender, child temperament, and teachers’ observed sensitivity when interacting with the children in their classroom. While Rappolt-Schlichtman and colleagues (2009) did not propose to examine HPA-axis reactivity, their results extend those reported by Lisonbee et al. (2008) to a younger sample of children in poverty, and are therefore included here: toddlers and preschoolers rated by their teachers as displaying higher levels of teacher-child conflict exhibited higher levels of cortisol when participating in small-group activities with their teacher and peers, but not when engaging in other classroom activities.

A final pair of studies examined the relation between non-parental caregiving behaviors and children’s basal cortisol levels using data collected during a large longitudinal study of young children (the Family Life Project). In the first of these studies, Berry and his colleagues found caregiver responsiveness, as assessed at 7, 15, 24, and 36 months of child age using the Home Observation for Measurement of the Environment (HOME; Caldwell & Bradley, 1984), was unrelated to children’s levels of basal cortisol at 48 months, after controlling for cortisol at 7 months. However, for children from households of low sociodemographic risk, more time in care predicted higher levels of basal cortisol, whereas for children from high-risk households, more time in care predicted lower levels of cortisol (Berry et al., 2014). In a second study, Berry and colleagues found that less responsive caregiving behaviors were associated with higher levels of basal cortisol at 7 months, but not at 15 or 24 months (Berry et al., 2016). It is important to note that unlike the studies of diurnal cortisol reviewed above, the studies conducted by Berry and colleagues address whether the relation between caregiving behaviors and cortisol transfers to the context of the child’s home. Perhaps the most surprising finding from these studies – that for children from high-risk households, more time in care was associated with lower cortisol levels – is broadly consistent with results reported by Rappolt-Schlichtman and colleagues (2009), in which cortisol levels among children in poverty decreased over the course the school day. Berry and Rappolt-Schlichtman offered similar explanations for their findings: placement in care may offer a reprieve for a child reared in a household of high risk. This explanation is supported by results reported by Chryssanthopoulou and colleagues (2005), who found that more time in care mitigated the association between children’s cortisol and maternal job stress and emotional exhaustion.

Discussion

In this paper we systematically reviewed the literature on the association between various dimensions of parental and non-parental caregiving behaviors and different aspects of HPA-axis activity in early childhood. This literature clearly demonstrates that parental and non-parental caregiving behaviors are associated with individual differences in multiple aspects HPA-axis activity. Although only two studies – both focused on parents – directly tested whether caregiving behaviors mediated the association between poverty and HPA-axis activity (Blair et al., 2011; Zalewski et al., 2012), this literature, together with the well-established relation between poverty and caregiving behaviors, suggests that parental and non-parental caregiving behaviors play an essential role in transmitting the effects of poverty to the activity of the HPA-axis in young children. Here we summarize and interpret the collective results of the studies included in our review, with an emphasis on how similar caregiving behaviors, observed among parents and early childhood educators, are associated with certain patterns of HPA-axis activity.

Various positive maternal caregiving behaviors were associated with individual differences in multiple aspects of HPA-axis activity across early childhood. For example, higher levels of maternal sensitivity were associated with lower baseline levels of cortisol (Berry et al., 2017; Blair et al., 2011; Grant et al., 2009; Kaplan et al., 2008; Laurent et al., 2016; Spangler et al., 1994), smaller increases in cortisol in response to challenge (Grant et al., 2009; Laurent et al., 2016; Spangler et al., 1994), and more rapid recovery to baseline levels of cortisol following a challenge (Albers et al., 2008; Laurent et al., 2016). Higher levels of other positive maternal caregiving behaviors also appear to down-regulate HPA-axis activity, as in the associations of engagement (Blair et al., 2008) and emotional availability (Bugental et al., 2003) with lower baseline levels of cortisol; comforting behaviors (Kiel & Kalomiris, 2016) and mindfulness (Laurent et al., 2017) with lower reactivity; involvement (Pendry & Adam, 2007) and warmth (Zalewski et al., 2012a, b) with lower diurnal cortisol; and emotional availability with lower bedtime (Philbrook et al., 2014) and overnight cortisol (Philbrook & Teti, 2016). Similar relations were observed between broad indices of positive maternal caregiving behavior and lower baseline cortisol (Blair et al., 2011) and reactivity (Martinez-Torteya et al., 2014; Thomas et al., 2017). For each maternal behavior or set thereof, results were generally consistent across samples of participants of varying socioeconomic composition (though see below for instances in which this was not the case). For example, the association between higher levels of maternal sensitivity and lower levels of baseline cortisol was observed in studies in which the majority of participants had high levels of education (Kaplan et al., 2008; Laurent et al., 2016), and in which the majority of participants were in poverty (Berry et al., 2017; Blair et al., 2011).

In contrast, higher levels of various negative maternal caregiving behaviors appear to up-regulate HPA-axis activity. Higher levels of intrusiveness are associated with higher levels of cortisol at baseline (Martinez-Torteya et al., 2015), larger elevations in cortisol in response to a range of stressors (Hutt et al., 2013; Martinez-Torteya et al., 2015; Taylor et al., 2013), and a prolonged recovery to baseline cortisol levels following challenge (Laurent et al., 2016). Other aspects of negative maternal caregiving behaviors – negative affect, punishment and conflict – are also associated with an increased likelihood of cortisol response to stress (Martinez-Torteya et al., 2014) and larger increases in cortisol (Hastings et al., 2011), as well as higher diurnal levels of cortisol (Hastings et al., 2011; Thompson et al., 2018; Zalewski et al., 2012a, b). The single study that examined paternal caregiving behaviors was consistent with these results (Mills-Koonce et al., 2011). As in the case of positive parental caregiving behaviors, these associations were generally consistent across samples that included substantial numbers of families in poverty (e.g., Martinez-Torteya et al., 2014) and in samples in which large majorities of families reported relatively high incomes (e.g., Hutt et al., 2013) or levels of education (e.g., Laurent et al., 2016)

A broadly similar pattern of associations was observed for non-parental caregiving behaviors and HPA-axis activity. Higher levels of sensitivity, whether measured directly (Groenevald et al., 2009) or by proxy through attachment security (Badanes et al., 2012), have been associated with lower diurnal levels of HPA-axis activity, as have stimulation and attention (analogous to parental engagement; Dettling et al., 2000a & b) and emotional support (analogous to emotional availability; Hatfield et al., 2013). In contrast, higher levels of intrusive or controlling behaviors have been linked to higher levels of diurnal cortisol (Gunnar et al., 2010). As in the case of parental caregiving behaviors, these associations were generally consistent across samples that included families from a range of socioeconomic backgrounds, with some samples including high proportions of families in poverty (e.g., Hatfield et al., 2013) and others low proportions (e.g., Dettling et al., 2000a & b). Although relatively few studies have examined the relation between non-parental caregiving behaviors and aspects of HPA-axis activity beyond diurnal function, those that did have yielded results that are broadly consistent with the literature on parental caregiving behaviors (Berry et al., 2014, 2016; Lisonbee et al., 2008). For example, Berry and colleagues (2016) found that less responsive non-parental caregiving behaviors were associated with higher levels of HPA-axis activity at baseline.

If positive caregiving behaviors (whether parental or non-parental) are associated with lower levels of HPA-axis activity, broadly defined, and negative caregiving behaviors are associated with higher levels of HPA-axis activity, what might be the implications of these alterations in activity for child development? Even when they fall short of the extreme represented by hyper-cortisolism, higher levels of cortisol in early childhood may adversely impact both short- and long-term development. Higher levels of cortisol can suppress aspects of immune function, rendering children more susceptible to certain chronic health conditions such as asthma (Segerstrom & Miller, 2004), and are associated with difficulties in self-regulation, such as internalizing behaviors (Gunnar, Tout, de Haan, Pierce, & Stanbury, 1997) and poorer performance on tasks of executive functions (Blair et al., 2011). Higher levels of cortisol in early childhood are associated with lower levels of subsequent cognitive functioning in the school years (Suor et al., 2015) and higher rates of cardiovascular disease in adulthood (Barr, 2017).

This does not mean that lower levels of cortisol are invariably adaptive – as noted above, hypocortisolism is also associated with poor outcomes (Blair et al., 2005; Gunnar & Vazquez, 2001; Heim et al., 2000) – but within the extremes represented by hypo- and hypercortisolism, lower levels of cortisol may promote development. This assertion is consistent with the multiple mechanisms that have evolved to keep levels of cortisol low in early childhood, including not only the aforementioned tendency towards reduced reactivity, but also declining levels of baseline cortisol (Hill-Soderlund et al., 2015; Ursache et al., 2014; Watamura, Donzella, Kertes, & Gunnar, 2004) and the emergence of diurnal rhythms that feature reductions in cortisol levels throughout the day (de Weerth et al., 2003; Matagos et al., 1998). More positive parental and non-parental caregiving behaviors appear to reinforce the tendency towards lower cortisol values, while more negative caregiving behaviors may undermine it.

It is important to note that there are some results among the studies reviewed here that contradict, or at least complicate, an account in which more positive caregiving behaviors are consistently associated with lower levels of cortisol. Although this link was generally observed across early childhood, in some longitudinal studies that included very young children, the association between parental caregiving behaviors and HPA-axis activity was not observed in the first three months of life. For example, the association between maternal emotional availability and levels of cortisol observed by Philbrook and colleagues at 3 months were not observed at 1 month (Philbrook et al., 2014).

Moreover, child factors other than age may also modulate the association between caregiving behaviors and HPA-axis activity. Most relevant to the present study is whether the child is from a household in poverty or poverty-associated risk. While associations between caregiving behaviors and HPA-axis activity were generally consistent across children drawn from samples of differing socioeconomic composition, there were exceptions. For example, the association between insecure maternal attachment classification and higher levels of baseline cortisol reported by Johnson and colleagues (2018) was only observed among children living in or near poverty, while the relation between over-reactive parenting and variability in diurnal cortisol was observed only for children of mothers who reported high levels of prenatal risk (Marceau et al., 2013). Other child factors may also play a role. For example, the association between positive maternal behaviors and smaller elevations in cortisol was not observed among children in some studies (Blair et al., 2015; DePasquale et al., 2018; Kertes et al., 2009). Rather, in these studies more positive maternal behaviors were associated with larger increases in cortisol in response to challenge, but in each case, this was among children with a particular trait or characteristic, such as emotional liability (Blair et al., 2015), social inhibition (Kertes et al., 2009), or a history of foster care (DePasquale et al., 2018).

However, the factor with the greatest impact on the associations between maternal caregiving behaviors and the HPA-axis response to challenge may be the nature of the challenge itself and, in particular, whether that challenge involves the mother. For example, Blair and colleagues (2008) found that higher levels of maternal engagement were associated with larger increases in cortisol in response to challenge among infants, but unlike in studies that used the SFP or SSP, in which positive maternal behaviors are generally associated with smaller increases in cortisol, the challenge used by Blair did not involve maternal interaction. A similar explanation may apply to results reported by Roque and colleagues (Roque et al., 2012). Even when maternal interaction occurs in the course of the challenge, associations between maternal behavior and reactivity may hinge upon whether that maternal behavior was observed in the context of the challenging task. For example, Sturge-Apple and colleagues (2012a) found that maternal emotional availability, assessed during free play and a compliance task, was associated with smaller increases in cortisol in response to the SSP. To the extent that maternal emotional availability as assessed during free play and compliance is indicative of maternal behavior more broadly, the episodes of maternal separation featured in the SSP may be less likely to evoke an increase in cortisol among children who are accustomed to a certain degree of maternal unavailability.

Directions for Future Research

At the outset of this paper we posed the question: what imbues poverty with the power to influence development? Drawing on ecological (Bronfenbrenner & Morris, 1998) and transactional (McLoyd, 1998) models, we asserted that the capacity of poverty to erode positive caregiving behaviors and promote negative ones (Holochwost et al., 2016; Popp et al., 2008), together with the fact that the activity of the HPA axis in early childhood is socially regulated (Hostinar et al., 2014), suggested that poverty’s influence on HPA-axis activity is based, in part, on its effects on caregiver behaviors. The literature reviewed here supports this account by demonstrating the associations between both parental and non-parental caregiving behaviors and young children’s HPA-axis activity. A clear next step for future research is to explicitly test whether specific dimensions of parental and non-parental caregiving behaviors mediate the relation between poverty and different aspects of HPA-axis activity. As noted above, two studies have examined this question with respect to parent behaviors (Blair et al., 2011; Zalewski et al., 2012a), but they necessarily confined their investigations to certain behaviors and select aspects of HPA-axis activity, and no investigation of which we are aware has examined whether and which non-parental caregiving behaviors might mediate the relation between poverty and the activity of the HPA axis, either independently or in conjunction with parental caregiving behaviors. One fruitful area for future research is how parental and non-parental caregiving behaviors jointly mediate the relation between poverty and HPA-axis activity, particularly if this examination were undertaken from a developmental perspective. For example, how might parental caregiving behaviors’ influence on HPA-axis activity in very early childhood impact children’s subsequent interactions with non-parental caregivers, and how might those interactions further affect children’s HPA-axis function?

However, future research should not focus exclusively on mediation. Caregiver bonds have been found to modulate HPA-axis reactivity to stress (Ahnert, Gunnar, Lamb, & Barthel, 2004; Gunnar et al., 1996; Nachmias et al., 1996), and therefore social interactions between caregivers and children may moderate the influence of poverty on HPA-axis activity. For example, more positive caregiving behaviors may buffer the influence of poverty, while more negative behaviors may exacerbate that influence (Gunnar et al., 2001; Szyf, Weaver, & Meaney 2007); to date these possibilities have rarely been assessed (cf., Fernald & Gunnar, 2009; Hibel et al., 2009). Moreover, when children are cared for by both parents and non-parental caregivers there is the possibility that these two forms of caregiving may jointly moderate the effects of poverty on children’s HPA-axis activity. Extending our previous example, poverty’s influence may be entirely mitigated if children experience positive caregiving behaviors from both their parental and non-parental caregivers. Similarly, that influence may be amplified if both parental and non-parental caregiving are negative. When patterns of parental and non-parental caregiving behaviors diverge – for example, if parental caregiving is positive and non-parental caregiving is negative – it raises additional questions, such as the relative strength of the moderating effects exerted by parental and non-parental caregiving behaviors and the extent to which a misalignment between caregiving received from parental and non-parental caregivers may itself influence HPA-axis activity. Although these questions are complex, they can be addressed in future research provided data regarding children’s HPA-axis activity in the context of parental and non-parental caregiving are collected along with data on the quality of the caregiving, something that has rarely been done to date.

Rather than adopting a perspective that treats mediation and moderation as mutually exclusive, future research should allow for the possibility that caregiving behavior functions as a mediator and a moderator, with its role changing contingent upon the specific parental or non-parental caregiver behavior(s) in question and the particular aspect of HPA-axis activity under investigation. For example, Zalewski and colleagues (2012a) found that among preschoolers maternal warmth mediated the association between poverty and blunted patterns of diurnal cortisol. This finding does not, however, preclude the possibility that other maternal caregiving behaviors moderate the effects of poverty on diurnal levels of cortisol, for example, or that maternal warmth moderates the effects of poverty on some other aspect of HPA-axis activity. Moreover, even when the behavior and aspect of HPA-axis activity under investigation are held constant, it may be possible that the role of caregiving behaviors will change. This change could depend upon levels of the caregiving behavior itself: at high levels, maternal warmth may buffer (i.e., moderate) the effects of poverty on HPA-axis activity, but low levels of maternal warmth may transmit (i.e., mediate) the deleterious influence of poverty to the development of the HPA axis. Finally, the role of caregiving behaviors may change in conjunction with some other variable (e.g., age). As noted above, all aspects of HPA-axis activity, from basal function to reactivity, are undergoing rapid consolidation in early childhood. All else being equal, the activity of the HPA-axis may be more open to the influence of poverty, whether direct or indirect via caregiving behaviors, when children are younger and those systems are less consolidated. However, as the process of consolidation progresses, the role of caregiving behaviors may shift towards moderation, such that HPA-axis activity is only open to the influence of certain configurations of poverty and caregiving behaviors.

Future research will also have to address certain issues of measurement. Whereas studies that used proxy measures of non-parental caregiving behavior (e.g., Badanes et al., 2012; Lisonbee et al., 2008) could allow for individual differences in the relations between a caregiver and different children, studies that have employed direct assessments of non-parental caregiving behavior have used measures (e.g., the CLASS) that do not allow for these differences. Priorities for future research therefore include adopting measures that can capture non-parental caregiving behavior as directed towards different children in the same classroom (e.g., the Individualized Classroom Assessment Scoring System, or inCLASS; Downer, Booren, Lima, Luckner, & Pianta, 2010) as well as the assessment of non-parental caregiving behaviors in early childhood education settings beyond center-based care. These priorities will require the adaptation of existing measures and, potentially, the creation of new ones that can yield valid indices of the quality of non-parental caregiving behaviors across settings (see Tonyan et al., 2017, for the challenges inherent in this work and current approaches to addressing them). These measures may also need to include the continuity of care as an essential aspect of the quality of non-parental caregiving behaviors. When explaining the finding that the number of non-parental caregivers was associated with higher levels of HPA-axis activity, Legendre (2003) argued that a larger number of caregivers may undermine the stability of the relationship between the child and any one caregiver, and other research indicates that stable relationships are necessary to down regulate HPA-axis activity (Hostinar et al., 2014), particularly in environments of elevated stress such as poverty (Garner et al., 2012). The fact that a particular non-parental caregiver exhibits high levels of positive behaviors and low levels of negative behaviors may be of negligible benefit to a child’s HPA-axis function if in short order that caregiver is gone from that child’s life. Multiple factors contribute to instability of caregiving arrangements of children in poverty, including staff turnover (Whitebrook et al., 2014), the use of multiple caregiving arrangements (Zaslow et al., 2006), and reliance on improvised caregiving arrangements that vary each day (Adams et al., 2007).

Conclusion

We have explored one way in which poverty may get under children’s skin (Lupien et al., 2001), reviewing the literature on the association between caregiving behaviors and young children’s HPA-axis activity. In doing so, we noted the potential for poverty to engender less positive and more negative caregiving behaviors (Holochwost et al., 2016; Popp et al., 2008). What we have not addressed, however, is how poverty may get under caregivers’ skin to alter their behaviors. Addressing this question is critically important if we hope to disrupt the associations between poverty and children’s HPA-axis activity. While there are many pathways through which poverty may impinge upon caregivers’ behavior, one mechanism that is central to this process is caregivers’ stress (McLoyd, 1998). This raises a question: can programs designed to alleviate stress among caregivers living in poverty serve as a means to mitigate poverty’s impact on children? Though not widely implemented, such “two-generation” programs do exist (National Research Council and Institute of Medicine, 2015, p. 167) and have been demonstrated to reduce stress among parental (Fisher, Gunnar, Dozier, Bruce, & Pears, 2006) and non-parental caregivers (Flook, Goldberg, Pinger, Bonus, & Davidson, 2013). Contributing to the design, evaluation, and improvement of such programs is one important way in which developmental scientists can better the lives of children and families, even as they wait for more ambitious, rational policies to alleviate poverty among parental (e.g., increasing the Earned Income Tax Credit) and non-parental caregivers (e.g., increasing wages) to become politically viable.

Highlights.

• Poverty is associated with HPA-axis activity in young children.

• Theory suggests caregiving behaviors mediate this association.

• We review the literature linking caregiving behaviors to HPA-axis activity.

• We focus on caregiving behaviors in the context of poverty.

• Our review includes both parental and non-parental caregiving behaviors.