Family risk, parental cortisol contagion, and parenting: A process-oriented approach to spillover

Abstract

This multi-method longitudinal study sought to investigate linkage in parental neuroendocrine functioning – indicated by cortisol – over two measurement occasions. In addition, we examined how parental cortisol linkage may operate as an intermediate factor in the cascade of contextual risks and parenting. Participants were 235 families with a young child (Mage = 33.56, 36.00 years for mothers and fathers respectively), who were followed for two annual measurement occasions. Parental cortisol linkage was measured around a laboratory conflict discussion task at both measurement occasions (i.e., pre-discussion, 20- and 40-minute post-discussion for each measurement occasion). Maternal and paternal parenting behavior was observed during a parent-child discipline discussion task. Findings indicated similar levels of cortisol linkage between parents over the two measurement occasions. Furthermore, cortisol linkage between parents operated as an intermediate factor between contextual risks and more compromised parenting behavior. That is, greater contextual risks, indicated by greater neighborhood risk and interparental conflict, were linked to greater cortisol linkage between parents over time, which was in turn linked to greater authoritarian parenting during parent-child interaction. Findings highlighted the importance of understanding physiological-linkage processes with respect to the impact of contextual risks on family functioning and may have crucial implications for clinical work.

Introduction

The bioecological theory of human development (Bronfenbrenner & Morris, 2006) highlights that individual behavior is shaped by multiple layers of contexts, including factors within families and proximal ecological contexts. Translated to the examination of the determinants of parenting behavior, the present study focused on contextual risks including difficulties within the family system and neighborhood risks. Specifically, prior theory and research have linked exposure to interparental conflict and neighborhood risks with compromised parenting behavior (e.g., Straus & Fauchier, 2007; Sturge-Apple et al., 2021). Yet, the specific processes through which neighborhood risk and interparental conflict may shape parenting are less understood. Drawing on the family risk framework (Repetti et al., 2002) and the allostatic load model (McEwen, 2002), we examined how linkage in parental neuroendocrinological functioning may operate as a potential intermediate factor in the cascade of contextual risks and parenting.

A process-oriented approach for neighborhood- and family-risk cascade

Previous empirical research has documented the effect of contextual risks within and outside the family context (e.g., interparental conflict: Krishnakumar & Buehler, 2000; Sturge-Apple et al., 2006; neighborhood risk: Chung & Steinberg, 2006; Cuellar et al., 2015; Barajas-Gonzalez & Brooks-Gunn, 2014) on parenting. Despite the comparatively well-documented direct associations, less research has adopted a process-oriented approach to understanding how contextual risks may be associated with compromised parenting behavior. Towards this, we drew on the family risk framework (Repetti et al., 2002), and allostatic load model (e.g., McEwen, 2002) to guide our investigation, both highlighting the role of alterations in physiological systems. More specifically, the family risk framework proposes that alterations in the physiological/neuroendocrine responses may operate as one of underlying mechanisms that explain associations between exposure to contextual and family adversity (e.g., family conflict and aggression) and adult functioning (e.g., mental and physical health; Repetti et al., 2002). Drawing on the allostatic load model, one may maintain physiological homeostasis via change (i.e., the process of allostasis, which characterizes the dynamic biological set-point of the body) when exposed to contextual risks. Yet, as characterized by the concept of allostatic load, repeated, severe and/or prolonged exposure to contextual risks may exceed one’s ability to regulate these changes, ultimately leading to “wear and tear” of the physiological system.

One particularly relevant system with respect to allostatic load is the hypothalamic-pituitary-adrenal (HPA) system, as it is one of the primary systems related to responding and coping with stressful events. When faced with stressors, the limbic system (e.g., hypothalamus) becomes activated and secretes the corticotropin-releasing hormone, which stimulates the secretion of adrenocorticotropin-releasing hormone by the pituitary gland and eventually results in the release of cortisol into the bloodstream by the adrenal cortex. The short-term increase in cortisol facilitates coping by mobilizing resources to process emotionally salient events, but long-term exposure to heightened cortisol levels may lead to “wear and tear” of the body (e.g., compromised immune functioning, impaired cognitive performance, high blood pressure; Gunnar & Vazquez, 2001; McEwen & Stellar, 1993).

Although prior research has linked exposure to family and neighborhood risks and altered neuroendocrinological activity in individuals specifically (e.g., Robinette et al., 2016; Sturge-Apple et al., 2009), limited attention has been dedicated to their impact on dyadic cortisol-linkage processes between individuals. To address this gap, here we focused on evaluating the cortisol linkage between parents during an interaction as a potential intermediate process that links contextual risks (i.e., interparental and neighborhood) to parenting behavior. Consistent with the previous literature (e.g., Timmons et al., 2015; Saxbe & Repetti, 2010), we use the term of “cortisol linkage” to characterize the linkage or covariation between two people in their adrenocortical functioning. We note, however, that different terms and operationalization has been adopted in existing literature to capture the linkage in cortisol levels or reactivity in adult couples (e.g., “coregulation of cortisol levels,” Saxbe & Repetti, 2010; “cortisol linkage,” Papp et al., 2013; Khaled et al., 2021; “cortisol synchrony,” Liu et al., 2013; Pauly et al., 2020; See more details in supplemental material, Table S1). Furthermore, existing work has adopted various approaches to assess cortisol (e.g., diurnal rhythm, cortisol reactivity around laboratory stressors), and utilized different adult participant groups (e.g., parents, young couples, expecting parents during pregnancy, older couples). Thus, given the variability of study design features in previous literature in this area, direct comparison of previous work to the current study is complicated. Accordingly, the current study design (i.e., testing cortisol around a laboratory conflict discussion and focusing on adult parents) tests very specific hypotheses that may not necessarily generalize to other participant groups or neuroendocrine assessment methods.

Our rationale to examine the cortisol linkage between parents was driven by Saxbe et al. (2020), who argued that the concepts of allostasis and allostatic load can be applied to social relationships. More specifically, social relationships may help regulate physiological arousal for each individual more efficiently (e.g., via instrumental and emotional support) and that theoretically, individuals within social groups may coordinate with each other physiologically that help to more effectively regulate each social partner’s arousal. Such physiological coordination has the flexibility to change when facing external stressors while maintaining a group-level homeostasis (i.e., social allostasis, e.g., an attuned spouse calming and comforting their agitated partner and bringing the overall arousal down). Yet, chronic, repeated, or severe exposure to stressors may result in “wear and tear” in the social relationships, and such “wear and tear” is defined as “social allostatic load” (i.e., parallel to “allostatic load” in the allostatic load model) that may damage the social relationship itself. One the physiological level, relationships with elevated social allostatic load may be manifested, for example, as strongly linked cortisol levels between partners, which might indicate a pattern of co-dysregulation that prevent individuals from adaptive physiological recovery (Mercado & Hibel, 2017; Saxbe et al., 2020). For instance, during conflict discussion, social relationships between parents who have been repeatedly exposed to contextual risks may have elevated social allostatic load, which could potentially be manifested as a stronger physiological linkage that could be driven by insufficient recovery after stress exposure, or parents being stuck in a loop of negative-affect escalation (Levenson & Gottman, 1983).

In turn, social relationships under such conditions might no longer be helpful in regulating arousal; in contrast, it might require the individuals to spend extra cognitive and/or regulatory resources to regulate themselves that may take a toll on their own adjustment and well-being (Saxbe et al., 2020). Given parenting can be a challenging process during which parents need to regulate themselves and provide appropriate socialization and caregiving for children (e.g., perceiving and interpreting children’s signals accurately, respond in an appropriate manner), parents’ depleted regulatory resources resulted from the elevated social allostatic load within the interparental relationship may hinder their abilities to parent their children in a sensitive, patient, and child-centered manner.

Taken together, our rationale to focus on cortisol linkage as a potential indirect factor that links family risks (i.e., neighborhood risk, interparental conflict) and parenting include the following points. First, the allostatic load (McEwen, 2002) model highlighted that altered HPA functioning may be one particular intermediate mechanism that accounts for the association between contextual risks and individual functioning. Second, the social allostatic load perspective specifically argued that contextual risks may result in strains in the social relationships, with elevated social allostatic load potentially indicated by stronger cortisol linkage among social partners (i.e., between the two parents). Third, prior research suggests that dyadic indicators of interparental relationship functioning may account for the link between family risks and parenting (e.g., insecure romantic attachment between parents, Davies et al., 2009). Examining cortisol linkage between parents as an indirect factor may help to explicate the intermediary role of interparental relationship functioning at the biological level of analysis and further contextualize previous research on psychological and behavioral levels of analysis. Taken together, a focus on cortisol linkage – an indicator for dyadic relationship (e.g., Timmons et al., 2015) – as a potential indirect factor for family-risk cascade is grounded in theory and prior research.

Interparental conflict, neighborhood risks, and cortisol linkage

A growing body of emprical research has evaluated the association between contextual risks (i.e., neighborhood risk and interparental conflict) and cortisol linkage in couples. Of the various approaches adopted by prior research (e.g., different adult groups, different approaches to measure cortisol), the present study utilizes a laboratory stressor design with parents, where cortisol linkage was conceptualized as the association in parents’ cortisol around the laboratory stressor. Given no prior research, to our knowledge, adopted the exact same study design and population, we drew on this body of work broadly as references even though they might be limited in guiding our hypotheses due to potentially different underlying physiological processes captured by cortisol reactivity vs. diurnal cortisol rhythms (e.g., Papp et al., 2012; Pauly et al., 2020).

Across this literature, inconsistent findings have emerged. First, a smaller body of work observed an association between adaptive and positive contextual factors and higher cortisol linkage. For example, Papp et al. (2012) examined linkage in married couples’ diurnal cortisol rhythms, finding that cortisol linkage was stronger for couples who spent more time physically together across the study period. Additionally, Pauly et al. (2020) documented associations between positive socioemotional interactions in older couples and greater diurnal cortisol linkage. Finally, Khaled et al. (2021) found that stronger cortisol linkage around a laboratory stressor among expecting couples was associated with fewer negative conflict behaviors and lower paternal postpartum depression.

In contrast, many other researchers have found that greater contextual risks (i.e., negative contextual factors) are associated with higher cortisol linkage in couples (i.e., adaptive, and positive contextual factors linked to lower cortisol linkage). These risk factors included greater marital strain and conflict (Liu et al., 2013), lower marital relationship satisfaction (Saxbe & Repetti, 2010), maternal-reported relationship aggression (Saxbe et al., 2015), and psychological stress (Braren et al., 2020). For instance, Saxbe and Repetti (2010) measured married couples’ diurnal cortisol levels and negative moods (e.g., angry, sad) four times per day (i.e., upon awakening, before lunch, right before leaving work, and before bed) for three consecutive days. Couples showed a positive cortisol linkage in their diurnal cortisol levels. Furthermore, couples reporting lower marital relationship satisfaction exhibited stronger cortisol linkage between each other. In addition, greater cortisol linkage was also linked to higher associations in couples’ negative moods. Taken together, prior research suggested that strains and conflict within the marital relationship may be associated with altered cortisol linkage between parents, although the direction of such association appears inconsistent (see Mercado & Hibel, 2017; Timmons et al., 2015 for hypothetical non-linear associations between environmental risks and physiological linkage).

In addition to interparental conflict, we were interested in examining the role of neighborhood risks given Saxbe et al. (2020) highlighted that social relationships can be perturbated by external stressors on a broader level outside the family. More specifically, this work suggests neighborhood risks may be a specific risk factor that associated with elevated social allostatic load in the family relationship. In turn, a stronger physiological linkage within the interparental relationship might reflect “co-dysregulation” between parents when their relationships are repeatedly and/or chronically affected by the risky neighborhood. Turning to empirical evidence, although there were no studies to our knowledge that directly assessed associations between neighborhood risk and couples’ cortisol linkage, our hypotheses were informed by prior research linking a) neighborhood risk and elevated allostatic load and b) neighborhood risk and parent psychological distress. Our decision to draw on these respective bodies of empirical work complements proposed theoretical frameworks in this area of research. According to the social determinants of health and environmental health promotion model (Schulz & Northridge, 2004), the effects of neighborhood risk on health may reflect systematic disinvestment in communities resulting in poor infrastructure and limited community resources, which in turn may manifest in increased stressors for community members, including physical and psychosocial stressors. Prior work (e.g., Robinette et al., 2016) has integrated the neighborhood-focused social determinants of health perspective with allostatic load models, positing that the physical and psychosocial stressors of risky neighborhoods may take a physiological toll on the human body. Results supported this perspective, finding that in Robinette’s adult sample, greater neighborhood risks were linked to higher overall allostatic load, which included cortisol level as one indicator for the risk within the HPA system (Robinette et al., 2016). These results align with the broader literature linking neighborhood risk to allostatic load (Brody et al., 2014; Robinette et al., 2016; Theall et al., 2012). Moreover, a recent review (Ribeiro et al., 2018) identified 14 studies examining the link between neighborhood socioeconomic deprivation and allostatic load, with 12 of 14 studies reporting significant associations between neighborhood deprivation and allostatic load.

Further indirect support for the hypothesized link between neighborhood risk and interparental cortisol linkage may be drawn from prior research on neighborhood risk and psychological distress (e.g., Giurgescu et al., 2012; Guterman et al., 2009; Jokela, 2020). More specifically, we expect such heightened psychological distress stemming from neighborhood risk may manifest in elevated cortisol linkage among couples by increasing negative-affect exchanges between parents that may increasingly tie their physiological arousal. This hypothesis is supported by previous research by Braren and colleagues (2020) which found a significant association between parent psychological distress and cortisol linkage between expecting mothers and fathers.

Cortisol linkage and parenting

A key aim of the present study is to evaluate parenting sequalae associated with cortisol linkage between parents, particularly parental discipline. Importantly, we focused on parental authoritarian discipline given two reasons. First, the links between contextual risks and parenting behavior are particularly strong within discipline contexts that may elevate authoritarian discipline (Straus & Fauchier, 2007; Sturge-Apple et al., 2021). Second, although no existing evidence has linked parents’ cortisol linkage with parenting behavior, prior literature has documented association between greater parental cortisol levels and reactivity with compromised parenting, with authoritarian discipline being one particular salient aspect highlighted by prior work. That is, parents with greater cortisol level/reactivity tended to show greater authoritarian discipline (e.g., Kiel & Buss, 2013; Martorell & Bugental, 2006; Mills-Koonce et al., 2009).

Conceptually, social allostatic load perspective (Saxbe et al., 2020) and negative-affect reciprocity model (Levenson & Gottman, 1983) both highlighted that couples with greater physiological linkage with each other might be stuck in the exchanges of negativity, having limited regulatory resource to parent their child in a child-centered manner. Yet, empirical evidence remains inconsistent in the form and direction of how cortisol linkage may be associated with individual outcomes. More specifically, there has been some emerging evidence on parental sequalae associated with linkage within other physiological systems (e.g., testosterone linkage linked to parent investment, Saxbe et al., 2017; association in electrodermal activity and thus sympathetic stress activity linked to parenting behavior, Papp et al., 2020), but research is limited in how cortisol linkage may be associated with parental functioning. For instance, in a study investigating broad sequalae of cortisol linkage, Pauly et al. (2021) examined diurnal cortisol linkage among older couples for seven consecutive days and assessed relationship satisfaction and physical health (indicated by cholesterol levels) over three annual measurement occasions. For older couples, although stronger cortisol linkage was linked to higher relationship satisfaction for wives, stronger cortisol linkage also forecasted poorer health for both husbands and wives over time across both general health and cholesterol levels specifically. Given the interrelated nature of relationships within family, the impact of cortisol linkage on individual functioning and relationship satisfaction may also extend to parenting (Chen et al., 2022). The direction of such association, however, might depend on the context (e.g., expecting parents during pregnancy vs. parents after children were born) and the specific neurobiological marker of interest (e.g., testosterone vs. cortisol). Furthermore, prior literature has been inconsistent in whether fathers or mothers might be more strongly affected by family risks. That is, some studies highlighted the greater potential variability in paternal parenting – compared to mothers – that paternal parenting might be more strongly shaped by family risks (e.g., Belsky et al., 1991; Szepsenwol et al., 2015). Others highlighted that mothers might be more sensitive and invested in close interpersonal relationships and family (Thompson & Walker, 1989), instead suggesting that maternal parenting might be more strongly affected by family risks (e.g., Sturge-Apple et al., 2009). Given these inconsistent prior research findings and limited work examining the association between physiological linkage and parenting, we did not have a specific hypothesis for the magnitude and direction of differences in how cortisol linkage might be linked to maternal vs. paternal parenting.

The present study

This study adopted a two-wave longitudinal design to examine how family risks cascade to parenting via interparental cortisol linkage over two years. We focused on the period when children were in early childhood (i.e., roughly ages three to four across the two measurement occasions) given this is a stage associated with elevated interparental conflict and parenting stress (Kwon et al., 2013; Williford et al., 2007). More specifically, children experience substantial changes in their cognitive and socioemotional abilities (e.g., gaining greater autonomy that may leads to noncompliance to parents’ request, developing self-regulating abilities but may still have limitations in regulating emotions and impulses, growing language skills but may still have difficulties to express themselves effectively). Although these abilities allow children to be more independent and autonomous, the growing abilities may pose challenges for parents who may still try to set boundaries and accommodating for children’s changing abilities. In addition, the transition to preschool for many children may add to the challenges as children and families navigating separation anxiety and adjusting to new social and academic environments.

Based on prior literature (e.g., Timmons et al., 2015), we hypothesize that parents will show linkage in their cortisol levels around the conflict discussion. Turning to the continuity and change in cortisol linkage over time, scarce literature has delved into this issue (DePasquale, 2020), with only one study to our knowledge that examined cortisol linkage between mother-child dyads over time (Hibel et al., 2015). Thus, we drew on this study to guide our present work, even though Hibel et al. (2015) focused on mother-child dyads whereas our study focused on parents. Within this study, Hibel and associates (2015) assessed mothers’ and young children’s cortisol levels around stressful laboratory tasks evoking anger, fear, and frustration for young children (e.g., arm restraint, scary mask) over three measurement occasions from early infancy to toddlerhood. Results indicated a significant linkage between maternal and child cortisol levels, such that when the mother had a higher cortisol level, the child did as well. In addition, although linkage in mother-child cortisol levels decreased after the stressful provocation, cortisol-linkage levels did not differ across the three measurement occasions. In other words, at different developmental stages of the child (i.e., infancy vs. toddlerhood), mothers and children seemed to maintain a similar level of cortisol linkage when coping with stressful laboratory challenges (directed to the child). Thus, even though the two studies focused on different participant groups and thus design and findings may not be comparable, here we hypothesize that cortisol linkage may show continuity over the two years. Nonetheless, given scarce empirical evidence, we considered this hypothesis preliminary and still examined the role of measurement occasion on cortisol linkage. Furthermore, we hypothesize that cortisol linkage over the two occasions (as we hypothesized continuity in cortisol linkage) may operate as an indirect factor that links greater family risks (i.e., neighborhood risks, interparental conflict) with greater authoritarian parenting at the second measurement occasion for both parents given theory and prior literature (e.g., Kiel & Buss, 2013; Levenson & Gottman, 1983; Mills-Koonce et al., 2009). Again, this hypothesis is preliminary given the same reason described above. This study contributes to literature in several ways. First, this is the first endeavor, to our knowledge, to examine the continuity and change in cortisol linkage between parents over time. Second, this study helps to elucidate the potential mechanisms through which contextual risks may be linked to parenting. Thus, this investigation is the first work to our knowledge to evaluate how cortisol linkage may be involved in these processes. Finally, another highlight of this study is the inclusion of both mothers and fathers, which helps to advance our understanding of the determinants of both maternal and paternal parenting. This contribution is notable, as fathers have typically been understudied in the parenting literature (Coley, 2001; Marsiglio et al., 2000).

Method

Participants

Participants were 235 families (mothers, fathers, and their approximately 3-year-old child) recruited from a mid-sized city in the Northeastern United States. Families were recruited from child-care centers, Head-Start programs, flyers and online posts on family-centered internet sites, and local community events. The original project from which this study drew data examined interparental relations and parenting, and the inclusion criteria were as follows: (a) having a target child who was at least three years old and both parental figures were at least 18 years old; (b) the three family members (i.e., two parental figures and the child) had been living in the same household for at least the prior year; (c) at least one of the parental figures is the biological parent to the target child, and two parental figures are of the opposite sex; (d) families members communicate in English, and (e) the target child does not have severe developmental disabilities. Wave 1 data collections took place between April 2017 and September 2018. Although the present study focused on parents (i.e., not children), we presented demographic information about the target children in this section to provide more detailed context of this study; also see more details of this project in our previous studies focusing on children (Li et al., 2021, 2022).

The average age of mothers, fathers, and their children were 33.56 (SD = 5.30), 36.00 (SD = 6.41), and 2.97 (SD = 0.38) years, respectively. 55.3% of the target children were female (N = 130). The median maternal and paternal education attainment was Associate’s degree and some college (less than two years), respectively. The median household income fell within the range of $55,000–74,999, although 25.5% families reported having a household income below $23,000. On average, 2.12 (SD = 0.52) and 2.65 (SD = 1.22) adults and children, respectively, live in the household. In addition, the three family members had been living together for 3.28 years (SD = 1.71). With regard to race and ethnicity, 62.6%, 55.3%, and 56.2% mothers, fathers, and child, respectively, were White, followed by 21.3%, 23.8%, and 21.3% mothers, fathers, and child being Black or African American (See more details in the supplemental material, Table S2).

A total of 218 families returned for the second family visit roughly a year after the first measurement occasion, corresponding to a retention rate of 92.8%. “Time-in” procedures were used to track when families were due for their second visit, and visits were scheduled as close as possible to 1-year from their first visit date based on family availability and responsiveness. Wave 2 data collections took place between June 2018 and August 2019 and the mean time elapsed between families’ Wave 1 and Wave 2 measurement occasions was 11.60 months (SD = 1.31). The mean age for mothers, fathers, and children were 34.54(SD = 5.29), 37.13(SD = 6.35), and 3.88(SD = 0.51) years, respectively. We obtained written consents from both parents before the family visit at each wave. The study protocol was reviewed and approved by the Institutional review board of the University of Rochester (Title of the study: Interparental Relationship and Parenting, case number: RSRB939).

Procedure

After the arrival of the center, experimenters gave each family a roughly ten-minute introduction of the study (e.g., overall goal of the study, tasks involved) before both parents completed the consent forms. Then, parents were led to quiet, separate rooms to complete several questionnaires for roughly ten minutes, which was followed by the collection of the baseline cortisol sample and the interparental conflict discussion task. Thus, each family member had at least 20 minutes of relatively quiet, non-stressful time before the baseline saliva sample was taken. The interparental conflict discussion task was the first interaction task of the visit. After the discussion, participants completed the dyadic parent-child interaction task in a counter-balanced manner (i.e., for a randomly selected half families, father-child interaction task proceed before the mother-child task, whereas for the other half families, it was the opposite order). Saliva samples for parents were collected by the experimenter simultaneously at three time points (i.e., baseline, and 20 minutes and 40 minutes following the ending of the 10-minute interparental conflict discussion task. Collection time for each cortisol sample were recorded (i.e., a total of six collection times: 3 times/measurement occasion * 2 occasions. See detailed study procedure timeline in Figure 1).

Figure 1. Timeline of the study procedure for each measurement occasion. Note. the study protocol was the same for these tasks across the first and second measurement occasions, so only one wave’s procedure was plotted here for parsimony.

Interparental conflict discussion (Wave 1 & 2)

Parents completed the 10-minute interparental discussion task (Sturge-Apple et al., 2009) at both measurement occasions. During the task, parents identified and discussed two topics that they commonly disagree about for ten minutes total, corresponding to five minutes per disagreement topic. The experimenter informed the parents that we were interested in how parents resolve disagreement. Thus, parents were given several minutes to (a) write down top three topics they commonly disagree about independently; and (b) take some brief time (i.e., a minute or two) to pick out two topics they would like to discuss during the interparental conflict discussion task. The experimenter also instructed parents to stay on topic during the entire task and knocked on the door from outside the visit room to signal the switch of topic midway through the task. Although this procedure does not involve standardized stressors, prior research has validated and widely used this paradigm to elicit conflict between parents, and that parents may show individual differences in cortisol reactivity to the discussion (Sturge-Apple et al., 2009). In addition, the advantage of using the conflict discussion paradigm lies in its greater ecological validity, as the task captures real-time conflict between parents.

Upon completion of the discussion, both parents rated the similarity of their discussion with the ones they usually had at home on a seven-point Likert scale (1 = “A lot more negative,” 4 = “very similar,” 7 = “A lot more positive”). Post-discussion surveys indicated that substantial number of parents rated the laboratory discussion to be similar to real-life interparental conflict in their family (i.e., 27.2% and 35.7% mothers and fathers reported the laboratory discussion to be “very similar” to the real-life discussion, respectively, 2.2% and 4.4% mothers and fathers endorsed “a little more negative,” respectively, and 37.1% and 28.2% mothers and fathers rated “a little more positive,” respectively).

Saliva collection (Wave 1 & 2)

At both measurement occasions, we collected saliva samples to measure parents’ cortisol levels. To limit the effects of diurnal cortisol rhythm (Stansbury & Gunnar, 1994), we collected saliva samples in the afternoon and early evening (1:00–6:00pm). Averaged collection time for the pre-task sample was 02:29 pm (SD = 2 hours, 13 mins). Before the family visits, experimenters informed the parents to follow their normal routines (i.e., time to sleep and wake-up), and avoid food, drink, and exercise at least one hour before arriving at the laboratory. We measured cortisol levels around the interparental discussion, which is a validated stressor that has been widely used in the previous literature (see above). As such, the first saliva sample was collected around 20 mins after the families arrived at the laboratory. This sample was used as the indicator for baseline cortisol levels. Given that cortisol reactivity generally peaks around 20-25 minutes after exposure to a stressor and recovers after around 40 minutes (Dickerson & Kemeny, 2004), we collected saliva samples at 20 and 40 minutes after the ending of the interparental conflict discussion task to index cortisol reactivity and recovery respectively. Regarding saliva collection, parents placed a cotton salivette into their mouth and chewed for one minute until the cotton was completely wet. According to prior literature, the salivette method provides reliable assessment of salivary cortisol levels (e.g., Schwartz et al., 1998). Salivary samples were immediately put back in the plastic tube and stored at −20°C until shipped on dry ice for assaying.

Discipline discussion (Wave 2)

At the second measurement occasion, each parent completed a separate, five-minute discipline discussion with the child. Before the task, the experimenter asked parents to write down a recent discipline issue involving the target child. Parents were then instructed to discuss the issue the way they normally would at home. Furthermore, parents were informed to stay on topic during the entire discussion period.

Measures

Neighborhood risk (Wave 1)

Neighborhood risks were assessed based on census data associated with the zip codes of family addresses at the first measurement occasion (http://www.city-data.com/). Zip codes of family addresses were gathered using maternal self-report. Greater neighborhood risks were indexed across a number of indicators, including higher unemployment rate, higher poverty rate within the neighborhood, higher percentage of renters, and higher population density. We standardized all four aspects and averaged them together to create an overall neighborhood risk score (e.g., Kirby et al., 2001).

Interparental conflict (Wave 1)

Trained staff observationally coded videotapes of the interparental conflict discussion task and rated the negativity and conflict using the System for Coding Interaction in Dyads (SCID, Malik & Lindahl, 2004) on a nine-point Likert scale (1= Not at all characteristic, 9 = Mainly characteristic). The two staff members were trained by a graduate student with substantial prior experiences in observational coding, who supervised the entire process. The two staff members double coded a few families each week, discussed discrepancies, and reached agreement during weekly coding meetings supervised by the graduate student with coding expertise. A random subset of tapes (20.0% of sample, n = 47) were double-coded and reached good reliability (Intraclass correlation [ICC] = 0.86). Greater negativity and conflict were reflected by higher levels of tension, frustration, irritation, and anger that were expressed in a non-aggressive or controlling manner (e.g., “you never listen to me”).

Cortisol (Wave 1 & 2)

Salivettes were stored at − 20°C until batch analysis at the end of data collection at the Laboratory of Biological Health Psychology (Brandeis University, Waltham, MA). Before assaying, saliva was centrifuged at 2000 g for 5 minutes. Cortisol was measured using a commercially available chemiluminescence immunoassay with detectable range of 0.3 to 86.4 nmol/l (CLIA; product #RE62119, IBL-International, Hamburg, Germany). Inter- and intra-assay coefficients of variation for assays were less than 10%. After removing outliers beyond ±3SD, we screened the cortisol data and found that none of the six samples of salivary cortisol were substantially deviant from normal distribution (range of skewness: [1.10, 1.90], so we did not perform any transformation (e.g., log-transformation) for the cortisol data.

Authoritarian parenting (Wave 2)

Authoritarian parenting was observationally coded during the parent-child discipline discussion task via the Caregiving Around Discipline System by two coding teams (CADS, Jones-Gordils et al., 2021). These two coding teams each had a master coder (i.e., a graduate student) who completed 100% of the coding for mother- and father-task videos, alongside two undergraduate research assistants who each double-coded 20% of the videos (i.e., one research assistant for mother-videos, and the other assistant for father-videos). The two research assistants were trained by the master coder, and separate weekly coding meetings (i.e., mother- vs. father- team) were held to address discrepancies in coding before reaching reliability. On a nine-point Likert-type scale (1= “Not characteristic at all,” 9= “Mainly characteristic”), authoritarian parenting refers to parents’ use of firm commands that require immediate compliance without question. Higher scores indicate parents frequently giving direct and consistent commands that leave no room for flexibility. Notably, this scale does not involve an affective component, so it captures parents firm control without hostility, belittling, or being excessively angry (ICC = 0.82 and 0.85 for mothers and fathers respectively). Thus, interpretation of the scale and comparison with other studies should be mindful that our authoritarian parenting measure did not involve the affective component (e.g., parents’ hostility, anger). Within our data (Table 1), parents showed comparable levels of authoritarian parenting (Mean = 2.60[mothers] and 2.53[fathers], t[196] = −0.14, p = .89). In addition, even though fathers seemed to have slightly more variability in their authoritarian parenting (SD = 2.08 [mothers]/2.20[fathers]), the variability in parenting between parents were fairly close.

Table 1. Descriptive information for primary study variables

Note. 1. These values mean the cortisol values for the corresponding individual at the corresponding collection time. For example, “Father Post-discussion Cortisol (Post-20 min)” means paternal cortisol value within the saliva sample collected 20 minutes after the ending of the interparental conflict discussion task. 2. The unit for cortisol is nmol/L. 3. These values (i.e., collection time) were the differences in saliva-sample-collection time vs. 00:00am the corresponding day, represented by hours (e.g., 2:12pm = 14 hours + 12min/60 min = 14.2 hours).

Data analysis plan

Given cortisol samples were nested within waves within person, we followed the structure of previous empirical work in examining cortisol linkage (e.g., Hibel et al., 2015; Khaled et al., 2021) and used a multilevel modeling framework to account for the nested structure of the data. More specifically, we adopted multilevel structural equation modeling (ML-SEM), which estimates cortisol linkage (i.e., random effect) in the momentary fluctuations as a latent variable and enables us to examine the antecedents and sequalae of cortisol linkage within the same model in one step. According to prior work, modeling via ML-SEM avoids the bias in model parameters caused by the unreliability of Empirical Bayes estimates (Liu et al., 2021). To do so, we first fitted null models for maternal and paternal cortisol data to estimate the level of cortisol linkage. Subsequently, we included the family-risk antecedents as predictors, and maternal and paternal parenting as outcomes of the latent cortisol linkage factor in ML-SEM. All analyses were performed in Mplus 8.0 (Muthén & Muthén, 1998-2011), and we used two-level models with Bayes estimator (Liu et al., 2021). Missing data were accounted for by full information maximum likelihood method in ML-SEM using all available data (see table 2 & 3 for numbers of observations in each model; Muthén & Muthén, 1998-2011).

Table 2. Estimate of cortisol linkage over time

More specifically, at each measurement occasion, each parent provided three cortisol assays: pre-task, 20-minutes post-task, and 40-minutes post-task (i.e., six total cortisol measurements possible). Thus, we treated cortisol measures as level one, and families as level two (i.e., we did not treat wave as a separate level for random effects given that we only had two waves). Following Hibel et al. (2015), we created three dummy codes to estimate the effects of measurement sample and waves. First, the pre-task dummy code estimated the differences in cortisol level before the task and 20 mins after the task. The pre-task sample was coded as 1, and the other two samples coded as 0. Second, the post-task dummy code estimated the differences in cortisol between the 20- and 40-minute post-task assays. As such, the 40-minute post-task occasion was coded as 1, and the other two occasions as 0. Third, the dummy code for wave was denoted as 0 vs. 1 for the first and second wave, respectively. We did not treat wave as a separate level given there were only two waves in our sample, and (a) treating wave as a level significantly increases the complexity of the model, (b) doing so potentially reduces the power given there were too few clusters on the corresponding levels (i.e., two cluster on the wave level, and three clusters on the cortisol sample per/wave level), and (c) there were clear distinguishability of the two waves (i.e., the first wave happened before the second wave) that can be sufficiently captured by a dummy code.

To estimate cortisol linkage between parents, and following prior literature (e.g., Hibel et al., 2015; Khaled et al., 2021; Helm et al., 2018), paternal cortisol on each time point was regressed on maternal cortisol, in addition to the three dummy codes for pre-task, post-task, and wave. To account for the effects of diurnal cortisol pattern, we included the time of collection (i.e., the differences in hours between the sample-collection time and 00:00:00) as two additional predictors (i.e., within-person centered time: estimating the effect of time on cortisol compared to the person-mean collection time, and between-person centered time: estimating the effects of averaged collection time on cortisol compared to the rest of the corresponding parents in the study sample). Furthermore, we also included several two-way interaction terms: (a) the pre-task-x-wave interaction, estimating the differences in changes of cortisol levels from pre-task to 20-minutes post-task between the two waves; (b) post-task-x-wave interaction, assessing whether the changes between 20- to 40-mins post-task differed across the two waves; (c) wave-x-cortisol interaction, pre-task-x-cortisol interaction, and post-task-x-cortisol interaction terms, measuring whether the cortisol-linkage level differ at the pre-task occasion, 40-mins post-task, and at the second wave, respectively. We ran models with mothers’ cortisol levels predicting fathers’ and vice versa. Although the substantive interpretation of both models indicate the level of cortisol linkage at the same occasion, we still ran the model both ways that was consistent with prior literature (e.g., Khaled et al., 2021).

Finally, we included the neighborhood risk and interparental conflict variables as predictors, and maternal and paternal parenting as outcomes associated with the latent cortisol linkage variable (See a conceptual diagram of ML-SEM in Figure 2). Following Preacher, Zyphur, and Zhang (2010), our predictors were included in level two (i.e., between-family level), the latent indirect variable (i.e., the mediator) – cortisol linkage – was on level two (i.e., between-family level, which was derived based on level-one associations in maternal and paternal cortisol), and our parenting outcomes were also on level two (i.e., between-family level). To avoid multicollinearity, we examined the role of neighborhood risks and interparental conflict within separate models. Yet, within each model, we included both maternal and paternal parenting. Indirect effects and the corresponding confidence intervals were estimated within Mplus via model constraints by creating the product of a (i.e., contextual risks → cortisol linkage) and b (i.e., cortisol linkage → authoritarian parenting) paths. According to Preacher et al. (2010), this method is a legitimate approach to obtain confidence intervals for indirect effects in ML-SEM in large samples like ours.

Figure 2. The conceptual diagram for the multilevel SEM model. Note. This diagram shows how we examined cortisol linkage as an indirect factor (denoted in the diagram as the coefficient ψ_i) between contextual risks and authoritarian parenting. Both contextual risks and authoritarian parenting were measured on level 2(i.e., between-family level), whereas the random effects of cortisol linkage (i.e., a latent factor, denoted as ψ_i on level 2, which was defined by level-1 associations in maternal & paternal cortisol). For easier interpretation, we did not plot the dummy codes included in the model (e.g., wave, pre-discussion, post-discussion) and the interaction terms. In addition, we only plotted one of the directions in level 1 (i.e., mother→ father cortisol).

Results

The descriptive information is presented in Table 1 (also see bivariate correlation for primary study variables in Supplemental material, Table S4). We now discuss findings from null models estimating cortisol linkage, followed by the role of environmental risks and parenting associated with cortisol linkage.

Cortisol linkage

In the null model with maternal cortisol predicting paternal cortisol (Table 2), we found significant effects of intercept, pre-conflict discussion, post-conflict discussion, wave, within-person time effects, between-person time effects, and cortisol linkage. These effects indicate that paternal cortisol level was different from zero (i.e., intercept). The significant pre-conflict discussion effect indicates that despite having stressful conflict discussion and the controlling for sample-collection times, the pre-discussion sample was still higher than the 20-min post-discussion (i.e., by 0.65). In addition, fathers showed a further drop in cortisol level from 20- to 40-mins post-discussion (i.e., by 0.28). Fathers exhibited a drop in cortisol level between the two waves as well (i.e., the second wave was 0.66 lower than the first wave). The two significant time effects both reflected that the later the collection time (compared to person-mean time, and compared to the averaged collection time with the other fathers in the sample), and the lower the cortisol levels. This was consistent with the pattern of diurnal cortisol rhythm, which gradually decreases during the day. Finally, paternal cortisol level was significantly associated with maternal cortisol level, indicating significant cortisol linkage.

Turning to the model with paternal cortisol predicting maternal cortisol, we found significant effects of intercept, pre-conflict discussion, post-conflict discussion, wave, within- and between-person time effects. In addition, we found significant pre-discussion-×-wave interaction, and post-discussion-×-wave interaction. Thus, mothers had a cortisol level greater than zero (i.e., intercept), showed a drop in cortisol level from pre- to 20-minutes post-conflict discussion (i.e., a drop by 0.66), and an additional drop from 20- to 40-mins post-conflict discussion (i.e., a drop by 0.47). Similar to fathers, maternal cortisol level was lower at the second measurement occasion compared to the first (i.e., −0.68 lower). The significant time effects also aligned with the diurnal cortisol rhythm, indicating lower cortisol levels with later collection time. The pre-discussion-x-wave interaction indicated that at the second wave, mothers showed a smaller drop in cortisol level from pre- to 20-mins post-conflict discussion (i.e., Wave 1: a drop of 0.66, Wave 2: a drop of 0.17 between the two cortisol samples). Furthermore, the post-discussion-x-wave interaction indicated that mothers exhibited smaller drop from the 20- to 40-mins post-conflict discussion (i.e., Wave 1: a drop of 0.47, Wave 2: a drop of 0.22 between the two cortisol samples). Finally, although the estimate of cortisol linkage was positive, this association did not achieve significance with paternal cortisol predicting maternal cortisol levels.

Contextual risks → cortisol linkage → parenting

As shown in Table 3, we first discuss findings involving neighborhood risks, followed by discussion on interparental conflict. For both directions of models (i.e., mothers’ predicting fathers’ cortisol, fathers’ predicting mothers’ cortisol), greater neighborhood risks were associated with higher cortisol linkage between parents modeled at wave one, which (i.e., the magnitude of cortisol linkage) did not differ across the two waves. Greater cortisol linkage, then, was associated with higher maternal and paternal authoritarian parenting during parent-child discipline discussion at the second measurement occasion. Tests for indirect effects indicate that the indirect pathways were all significant. That is, greater neighborhood risks were associated with greater cortisol linkage between parents, which was in turn associated with higher maternal and paternal authoritarian parenting one year later (Table 4). Furthermore, because we did not find a significant wave-x-cortisol interaction, we did not test the indirect pathways of neighborhood risks to authoritarian parenting over changes in cortisol linkage over the two occasions.

Table 3. Family risk factors and parenting behavior associated with cortisol linkage

Note. All SDs are posterior standard deviation.

Table 4. Estimates for indirect effects

Turning to the role of interparental conflict, for both modeling directions, greater interparental conflict was linked to higher cortisol linkage between parents again modeled at waves one and did not differ across the two waves. In addition, although greater cortisol linkage seemed to be linked to greater authoritarian parenting for both parents, the association between cortisol linkage and parenting was only significant for paternal parenting but not maternal parenting. Tests for indirect effects indicated that the indirect effects were significant for paternal parenting. That is, greater interparental conflict was associated with greater cortisol linkage between parents during conflict discussion, which was then associated with higher paternal authoritarian parenting (Table 4). Similar to the previous findings, we did not examine the indirect pathway linking interparental conflict, changes in cortisol linkage between the two measurement occasions, and authoritarian parenting due to non-significant wave-x-cortisol interactions.

Discussion

This study investigated parents’ cortisol linkage around a conflict discussion over two measurement occasions. We found a similar level of cortisol linkage over time, and that cortisol linkage may operate as an indirect factor between neighborhood risk and interparental conflict in promoting authoritarian parenting. Findings contribute to the literature by documenting continuity in cortisol linkage between parents over time, indicated by a lack of significant differences in the absolute values of cortisol-linkage over the two occasions. Furthermore, our results help to gain a greater understanding of the potential intermediate processes through which family risks may affect parenting. We discuss these findings below.

For both parents, we observed a significant drop in cortisol levels from pre- to 20-minutes post-discussion, and a further drop between 20- to 40- minutes after the discussion. In addition, the effects of cortisol collection times emerged, such that the later the collection times, the lower the cortisol levels. These findings proved generally consistent with the diurnal rhythms of cortisol (e.g., DeSantis et al., 2007; Gunnar & Vazquez, 2001; Liu et al., 2013). That is, cortisol, after peaking shortly after waking up, shows a gradual decline during the day. Thus, even though we restricted our family visits to a shorter time window mostly in the afternoon following well-established protocol (e.g., Davies et al., 2007) and controlled for sample-collection times, we still observed a decline in cortisol levels. Although our use of a well-established interparental discussion task did not elicit an increase in cortisol for the sample as a whole (e.g., Davies et al., 2016; Sturge-Apple et al., 2009), this is consistent with several prior research in adult participants (e.g., Flanagan et al., 2018; Hibel & Mercado, 2019; Powers et al., 2006). Moreover, consistent with the aims of this paper, there were substantial individual differences in intraindividual changes in cortisol in response to the task (See supplemental material, Table S3).

Inspecting the cortisol levels across the two waves, both parents manifested a decrease in the overall cortisol levels from the first to the second measurement occasion. After examining sample-collection time across the two waves, though, it was not the case that the differences in cortisol levels across waves were attributed to protocol/timing differences (pre-discussion saliva sample-collection time: mean differences between waves = −0.23, t = −1.35, p = .18; post-20 min collection time: mean differences = −0.22, t = −1.34, p = .18; post-40 min collection time: mean differences = −0.20, t = −1.19, p = .23). We speculate the differences in cortisol levels across two waves might be attributed to parents’ reduced overall stress between the two annual measurement occasions. Yet, given that we did not measure parents’ subjective stress levels at either measurement occasion, it is impossible to determine whether this is the case. One additional potential explanation is that parents, after visiting the laboratory on the first measurement occasion, became less stressful and more comfortable with the interaction task. In addition to the main effect of measurement occasion, we found significant wave-x-task-occasion interactions for mothers, such that mothers demonstrated a smaller dip in cortisol values from pre- to 20-minutes post discussion, and 20- vs. 40-minutes post discussion. These less pronounced drops in cortisol levels across task occasions at the second wave might be attributed to lower overall levels of cortisol that mothers had at the second wave.

Turning to linkage between parents during the interparental conflict discussion, we found a relatively small, and yet significant cortisol linkage between parents when paternal cortisol was predicted by maternal cortisol in the model. When maternal cortisol was predicted by paternal cortisol, although the estimated linkage was positive, it did not achieve statistical significance. Despite that, for both models, cortisol linkage showed significant individual variances between families, indicating that families differed from each other in the magnitude of cortisol linkage. The way that we modeled cortisol linkage aligned with prior research (e.g., Braren et al., 2020; Khaled et al., 2021) in including the prediction for both directions (i.e., mother → father, father → mother). Given that the cortisol linkage we documented were based on concurrent associations (i.e., we did not control cortisol levels assessed at prior time points), we are not well positioned to interpret the directionality of our findings here. Taken together, our findings generally aligned with prior research in revealing a linkage in parents’ cortisol levels (e.g., Liu et al., 2013; Saxbe & Repetti, 2010; Timmons et al., 2015). Such linkage in cortisol levels might be attributed to exposure to common stress, shared feelings of negative affect, and parents’ reciprocal influences on each other. Yet, the somewhat weaker linkage in our study may be caused by the differences in study design between our study and other published research. Namely, whereas we assessed cortisol linkage around a laboratory stressor over shorter time (i.e., within an hour), the majority of the prior research revealing a positive cortisol linkage investigated diurnal cortisol levels assessed multiple times a day over a multiple days (e.g., Meyer & Sledge, 2020; Timmons et al., 2015). Thus, our findings of cortisol linkage needed to be interpreted within the context of task (i.e., around the laboratory stressor of interparental conflict discussion), which did not evoke significant cortisol increases on average within the entire sample. Furthermore, the fact that our evidence for cortisol linkage was somewhat weak (i.e., a significant linkage when paternal cortisol was predicted by maternal cortisol, but not vice versa) requires replication by future studies.

The null findings for wave by cortisol interaction for both mothers and fathers indicated that the absolute magnitudes/levels of cortisol linkage between parents did not differ between the two measurement occasions. This finding proved consistent with the only prior study to our knowledge that examined the continuity in mother-child cortisol linkage, which also demonstrated no significant differences in the levels of cortisol linkage over time (Hibel et al., 2015). Families may maintain continuity unless disrupted by external perturbations (e.g., parent psychopathology, natural disaster; Davies et al., 2004; Minuchin, 1985), and thus the continuity in cortisol linkage that we documented here may be accounted for by a lack of significant perturbations to the family members between the two measurement occasions. Despite this point, here we were limited by only two measurement occasions, and thus the most meaningful contributions of the current study to the literature concern understanding cortisol linkage within waves rather than robustly characterizing continuity and change in cortisol linkage over several waves of measurement. To gain further understanding of this issue, future research should examine the continuity and change in physiological linkage among more measurement occasions and/or around external perturbations.

Turning to the association between contextual risks and cortisol linkage, consistent with prior research (e.g., Braren et al., 2020; Liu et al., 2013; Saxbe et al., 2015; Saxbe & Repetti, 2010), we found that greater contextual risks, either in the form of neighborhood risk or interparental conflict, was associated with higher cortisol linkage between parents during the conflict discussion. Our findings appear consistent with the allostatic load model (McEwen & Stellar, 1993), the social allostatic load perspective (Saxbe et al., 2020), and the negative-affect reciprocity model (Levenson & Gottman, 1983). That is, couples with greater contextual risks may show altered neurophysiological functioning that may be manifested within the dyadic interaction and physiological linkage. For instance, the social allostatic load perspective (Saxbe et al., 2020) highlighted that exposure to the contextual risks may pose repeated hits within the social groups, increasing the likelihood for conflict and discord escalation (e.g., a loop of negative-affect exchanges or shared stressful feelings), and delaying the recovery after the exposure to stress. Despite the possibilities, we did not directly measure parents’ subjective feeling of distress or conflict escalation or measure physiological activities on a more intensive manner (e.g., real-time measure that may happen on a second-to-second basis), and thus were not able to directly test these possibilities. It is also important to notice that our measure of parental cortisol linkage took place during the interparental conflict discussion that may naturally involve more conflictual exchanges between parents to observe the physiological linkage.

Notably, contradictory to what we found here, some prior research linked family risks with lower cortisol linkage (e.g., Khaled et al., 2021). Given these inconsistencies in the prior literature, we propose a similar interpretation advanced by prior researchers (e.g., Mercado & Hibel, 2017; Saxbe et al., 2020; Timmons et al., 2015) that a more specified and nuanced model for the meaning of cortisol linkage needs to be developed that accounts for the task context (e.g., stressful conflict discussion vs. support task), the specific physiological system of interest (e.g., HPA, sympathetic and parasympathetic nervous system activity), and the actual level of the physiological linkage (e.g., whether there is an “optimal” level of linkage, with higher and lower linkage being maladaptive). Yet, given our findings (i.e., that greater cortisol linkage was associated with greater family risk) and the goal of the study was to examine cortisol linkage as an indirect factor, we did not examine the non-linear association proposed by the theory pieces.

Turning to the indirect effects, we found that neighborhood risks were linked to higher cortisol linkage across the two waves, which was then associated with higher maternal and paternal authoritarian parenting. The greater interparental conflict was also associated with higher cortisol linkage, yet the indirect pathway only proved significant in forecasting paternal authoritarian parenting behavior. These findings proved generally consistent with theory and prior research (e.g., Sturge-Apple et al., 2006; 2009) such that contextual risks may promote parenting behavior that demands immediate obedience and compliance without questioning. Consistent with family risk framework and allostatic load model (Repetti et al., 2002; McEwen & Stellar, 1993), our results supported the notion that alterations in the physiological system, as manifested by stronger dyadic cortisol linkage between parents, may help to account for the effects of contextual-risk cascades. That is, parents who are stressed by risky neighborhoods or anger and conflict in the interparental relationship be more easily entrenched in negative-affect exchanges with one another and thus may become increasingly tied in their physiological states. This heightened tie in physiological states, manifested in higher cortisol linkage, may exceed parents’ ability to contain the stress within themselves, and remain patient and calm during discipline. These parents, in turn, may use discipline tactics that require immediate compliance (i.e., being authoritarian) due to it being less taxing or demanding on them as a parent. Further analyses that included additional covariates (i.e., child age, parental education) indicate that our findings for these indirect effects were generally robust to the controlling of additional demographic covariates (See supplemental material, Table S5-S8). As an additional note, even though authoritarian parenting has been regarded as a maladaptive form of parenting (e.g., Maccoby, 1994), recent research has highlighted that the meaning and implications of authoritarian parenting should be considered within the context (e.g., Deater-Deckard & Dodge, 1997; Sturge-Apple et al., 2022). Thus, authoritarian parenting might not universally undermine a child’s functioning; rather, authoritarian parenting style might be adaptive in some socioeconomic contexts and have certain advantages. For instance, prior research suggested that authoritarian parenting may play a protective role for children within unsafe neighborhoods, because more restrictive, and no-nonsense parenting may more effectively protect children from harm within their risky contexts (e.g., Deater-Deckard & Dodge, 1997; Bornstein & Zlotnik, 2009).

Finally, although higher cortisol linkage seemed to promote authoritarian parenting for both fathers and mothers, the effect in fathers was more robust. This may be accounted for by two reasons. First, compared to mothers who are usually the primary caregiver and thus may show less overall variability in parenting, paternal parenting may have higher individual differences and are more susceptible to the influences of other risk factors (e.g., Belsky et al., 1991; Szepsenwol et al., 2015). Within our data, fathers seemed to have a slightly higher variability in their authoritarian parenting compared to mothers, even though these values were fairly close. In addition, mothers usually operate as the gatekeeper of families who may either encourage or impede paternal involvement with child caregiving (e.g., Schoppe-Sullivan et al., 2015). Thus, paternal parenting behavior might be shaped by their interaction with mothers, including cortisol linkage during mother-father interactions.

Several limitations are worth mentioning before drawing conclusions. First, the present sample consisted mostly of middle- to low-SES, two-parent families, thus somewhat limiting generalization (see supplemental material Table S9 and 10 for linkage between family risk, cortisol linkage, and parents’ authoritarian parenting tested separately by parents’ ethnic-minority status). Second, although prior research utilized various methods to assess cortisol linkage (e.g., diurnal cortisol, around a stressor), we only measured parents’ cortisol several times around the conflict discussion during laboratory visits (see prior example: Khaled et al., 2021). Future research is thus encouraged to compare the similarity and differences in cortisol linkage assessed via different methods. In addition, although the interparental conflict discussion task has better ecological validity compared to a standardized laboratory stressor, the former does not present stressor at a standardized time and intensity, thus participants may show variabilities in their true peak time of cortisol responses. Towards building on this point, the interparental conflict discussion did not evoke a significant increase in cortisol levels in the whole sample (despite substantial individual differences), and thus all findings should be interpreted within the context of null overall cortisol reactivity. Third, although we aimed to examine continuity and change in cortisol linkage, we are limited to only two-time points spaced one year apart. This might be one potential reason that we did not find significant differences in cortisol linkage over time. Future research is thus encouraged to investigate this issue with more measurement occasions, across longer developmental periods, and around significant external perturbations (e.g., around children’s transition to puberty). Furthermore, the present study only had six cortisol samples (i.e., three cortisol samples per wave * two waves). Thus, the number of cortisol samples might be limited even though our sample size was acceptable in the multilevel frameworks (e.g., Hox et al., 2010; Preacher et al., 2010) and that such sample size is comparable to previous studies (e.g., six cortisol samples * 103 families of mothers, fathers, and adolescents; Saxbe et al. [2014]; six cortisol samples *82 couples, Khaled et al., [2021]; three cortisol samples/per day * 2 consecutive days * 385 couples, Braren et al., [2020]). Finally, building on the prior point, we did not examine the lagged effect of cortisol (e.g., cortisol [t-1] → cortisol[t]) given the number of our cortisol sample, which is a common scenario in this line of research (e.g., Saxbe & Repetti, 2010; Papp et al., 2013; Khaled et al., 2021; Liu et al., 2013; Pauly et al., 2020). Taken together, we suggest future research to include more cortisol samples to examine cortisol linkage. Relatedly, a design with more cortisol samples may allow for simultaneously consideration of cortisol levels, cortisol reactivity, and cortisol linkage between parents (whether its modeled only over several baseline samples, or across multiple samples over different occasions), and how these biological processes may be related to family risk and parenting As an additional note, we were interested in evaluating how contextual risks (e.g., interparental conflict) may be associated with cortisol linkage over the two occasions. We suggest future studies may examine the transaction between contextual risks and cortisol linkage using repeated measures for both factors (i.e., interparental conflict & cortisol linkage both measured across multiple time occasions). Thus, a more comprehensive examination is warranted that examine how couple dynamics on the behavioral level, cortisol levels and reactivities, and cortisol linkage might transactionally shape one another over time, and the role that the change in cortisol linkage over time might play in these relations. Fourth, although parent-child discipline discussion was not designed to evoke parents’ physiological stress, and that timing of the tasks were carefully designed so that discipline discussion did not confound the final cortisol sample (e.g., usually the 40-minutes post-interparental conflict discussion task cortisol sample was collected right after the discipline discussion for one of the parents, thus the timing was too short for cortisol reactivity in the discipline discussion to manifest), we did not test whether the final cortisol sample can be confounded by the parent-child interaction tasks. Finally, this paper only focused on one indicator for the HPA system functioning (i.e., cortisol), therefore, future research is encouraged to explore physiological linkage while considering multiple systems simultaneously to gain a more thorough understanding of the collective physiological functioning.

Despite these limitations, this study is the first endeavor to our knowledge to illuminate the continuity of cortisol linkage between parents over time. Cortisol linkage, in turn, operated as an indirect factor between neighborhood and family risks and elevated authoritarian parenting. Findings contribute to the growing literature in understanding antecedents of maternal and paternal authoritarian parenting, highlighting the importance of physiological-linkage processes with respect to the impact of contextual risks on family functioning. Findings also have crucial implications for clinical and social work as well as intervention in mitigating the impact of contextual risks on the family. For example, by identifying physiological linkage functioning as an indirect factor for contextual risks, our findings highlighted potential target for therapy and intervention, which could help parents de-escalate during conflict and thus reduce their physiological linkage and the related parenting sequelae. Furthermore, future work that may non-invasively track physiological linkage and provide real-time biofeedback within family interactions would potentially have therapeutic implications in improving parenting quality and family adjustment.