Parent Cortisol and Family Relatedness Predict Anxious Behavior in Emerging Adults

Abstract

Emerging adult cortisol response during family interaction predicts change in EA anxious behavior during the transition to college (Gans & Johnson, in press). In the present study, we take an additional step toward integrating family systems research and physiology by including assessment of parent physiology. We collect salivary cortisol from parents and emerging-adults during triadic family interaction. Emerging adults (N = 101) between the ages of 17 and 19 were assessed at three time points across their first college year: the summer before college, fall and spring semesters. Two parents accompanied the emerging adult child to the summer assessment; all family members provided four saliva samples each at 20-minute intervals. Later assessments of emerging adults included measures of internalizing behaviors. Parents’ cortisol secretion patterns during family interaction predict their emerging adult child’s cortisol secretion pattern, parent perceptions of the family environment, and emerging adult children’s internalizing behavior during the college transition. Different patterns of results emerged for mothers’ and fathers’ cortisol response to family interaction, and for families with sons or with daughters. The approach taken by this study provides a first step toward understanding how interrelationships among elements of physiology and family functioning contribute to adjustment during major life transitions.

Our approach to understanding the relationship between physiology and emotional adjustment examines stress physiology within a triadic family context. Previously, we have found students’ cortisol response during family interaction predicts change in their anxious behavior during the transition to college. The pattern of results differs between men and women, and is moderated by family cohesion (Gans & Johnson, in press). In the present study, we expand this work to include assessment of parent physiology by using the opportunity of observing triadic family interaction to collect salivary cortisol from parents and emerging-adults at the start of a normative developmental transition (i.e., the transition to college). While the family literature largely views cortisol secretion patterns as a result of family dynamics, and the physiological literature views it as a cause of successful adaptation to challenging circumstances, we argue that cortisol secretion is an important component of family dynamics. Such an approach is supported by studies of married couples; autonomic nervous system reactivity is viewed as a component of couple interaction, and is linked during conflict in couples with low marital satisfaction (e.g., Levenson & Gottman, 1983).

Up to the present, what research exists on child cortisol and family dynamics has focused mainly on insider perspectives (i.e., self-report) of early and current family functioning, and few have examined cortisol levels from both parents during a task (Saxbe, Margolin, Shapiro, Ramos, Rodriguez, & Iturralde, 2014; Stenius, Theorell, Lilja, Scheynius, Alm et al., 2008). Moreover, no study has investigated linkages between parent cortisol response and children’s adjustment to normative developmental change. Although few in number, some empirical study finds linkages between parent and child cortisol response and between maternal cortisol response and parenting behavior (Hibel, Granger, Blair, & Cox, 2001; Kiel & Buss, 2013; Saxbe et al., 2014; Stenius et al., 2008). Mothers’ and fathers’ trait cortisol response is correlated with their infant’s cortisol response (Stenius et al., 2008). Whereas strong correlations were found between mothers’ and infants’ cortisol levels across the day, a less robust relationship was found between fathers’ and infants’ cortisol response, and only during afternoon and evening assessments. In one of the few studies to assess cortisol response during triadic family interaction, Saxbe and colleagues (2014) find positive associations among cortisol levels of mothers, fathers, and adolescents. Mothers’ cortisol has also been found to predict parenting behavior with toddlers (Kiel & Buss, 2013); mothers with high cortisol levels were found to be intrusive when interacting with their two-year-old children. Hibel and colleagues (2011) find sensitive maternal behavior to moderate the relationship between mother report of intimate partner violence and her toddler’s cortisol response to a challenge task.

Gordis and colleagues (2010) consider triadic concordance in salivary alpha amylase during a family conflict discussion, and find evidence mainly for dyadic concordance that is moderately dependent upon family cohesiveness. Mothers and their adolescent children were found to have concordant salivary alpha amylase responses at baseline and after the conflict task, particularly in families with no reported interparental aggression. Concordance between fathers and children was moderated by interparental aggression, so that families reporting higher levels of aggression demonstrated greater father-child concordance. These results support our hypothesis that parent physiology can be predictive of child psychosocial outcomes.

The current study builds on the existing literature by using outsider observations of family relatedness as well as insider perspectives of the family environment (i.e., parent self-report), and by using the opportunity of observing family interactions to collect salivary cortisol from parents as well as their offspring. This approach allows us to utilize cortisol in a novel way, to build an understanding of physiology at the level of the family and to begin to understand the ways in which parent cortisol may be a biomarker for adjustment during the transition to college. More specifically, we test the following three hypotheses: (1) parents’ individual cortisol response to family interaction will predict their child’s cortisol response to the same family interaction; (2) parents’ individual cortisol response to family interaction will be linked to family functioning, and (3) parents’ individual cortisol response to family interaction will predict changes in their child’s depressed and anxious behavior across the first college year.

Emerging Adult Adjustment

While it is has been established that early family functioning plays a role in both emotional outcomes and in physiological functioning, no study has observed the interaction among these variables across a relevant developmental time span. We propose to investigate changing cortisol responses within the family context during emerging adulthood, a time marked by increasing independence and rapid-fire social and emotional challenges (Arnett, 2000).

The transition to college has provided an opportunity to observe family functioning within an ecologically relevant context. This has allowed for the elucidation of the psychobiological mechanisms contributing to the emergence of the internalizing behaviors in emerging adults (Gans & Johnson, in press). The transition to college provides a somewhat unique window into understanding how cortisol patterns are linked to family functioning and individual adjustment as emerging adults transition away from the family home and toward a semi-autonomous independent college experience.

About one third of young adults begin the developmental period of Emerging Adulthood with the transition to college (Arnett, 2000). There is a growing literature identifying correlates of emotional adjustment during the college transition (Johnson, Gans, Kerr, & Deegan, 2008; Johnson, Gans, Kerr, & LaValle, 2010). For example, Johnson and colleagues (2010) find that when new college students reported more cohesion and expressiveness in their family, they also reported high levels of emotional adjustment to college. Similarly, emerging adults from families with high levels of conflict reported difficulty with their emotional adjustment during the transition to college.

What makes this understanding so urgent are statistics related to the emergence of pathology during the college years. Ten percent of college students are diagnosed with depression, and between 1.7 and 3.3 percent of college men and women receive therapy for internalizing problems while attending college (American College Health Association, 2007). One-month prevalence rates for depression are highest for adolescents and emerging adults (ages 15 to 24 years) than for any other time of life (see Berry, 2004). Similarly, many anxiety disorders (e.g., panic disorder, generalized anxiety disorder, agoraphobia) are common in late adolescents and early adulthood (Anxiety Disorders Association of America, 2007). The relationship between family functioning and change in anxious behavior across the college transition year may depend on the new college students’ cortisol response to family interaction (Gans & Johnson, in press).

With few exceptions, previous research examining the transition to college is limited in several important ways. First, most college adjustment research relies on data collected after students have already begun college, thereby missing participants’ actual college transition. Second, few studies take a longitudinal and prospective approach to examining college adjustment. Most research in the area of emerging adults’ adjustment to college focuses exclusively on students' initial adjustment to their first semester, foreclosing the opportunity to investigate correlates to changes in student adjustment over time. Third, research on emerging adult development in general rarely assesses family functioning in family triads or in a way that includes physiology. The present study, therefore, is unique in its prospective, longitudinal approach to examining biopsychological predictors of college adjustment assessed within a triadic family context prior to emerging adults’ arrival on campus and throughout the first college year.

Method

The present study uses a multi-method approach, including observational, physiological, and self-report data, to examine the role of individual physiological response within a family context in predicting later adjustment during a normative developmental transition.

Participants

Three cohorts of a total of 101 first-year undergraduate students (37 male, 64 female) at a public state university in the North Eastern United States participated in the present project. After accepting admission to the university, new college students were contacted by mail requesting that they participate in a project studying new students’ adjustment to the first year of college. Inclusion criteria required student participants to be between the ages of 17 and 19, to be starting college for the first time, and to have two caregivers who also agreed to participate in the study. In an effort to obtain some homogeneity in our definition of family among our sample, we defined caregivers as two adults with whom the new student participant has been living for the last 5 years prior to starting college. Mean age of student participants was 17.85 years (SD = 0.45 years), 89% were Caucasian, 5% self-identified as African American, 5% self-identified as multi-racial, and 1% self-identified as Latina/o. Mean age of mothers was 49.05 years (SD = 4.35 years) and mean age of fathers was 51 years (SD = 5.14 years). Attrition between initial study participation (pre-college assessment) and the final follow-up assessment (spring semester) resulted in complete data for 83 student participants (25 male, 58 female). There were no significant mean differences in pre-college study variables for participants based on study completion.

Procedure

Student participants came to the Whole Family Biopsychology laboratory for three assessment appointments over the course of their first year in college: (1) Pre-College Assessment (during the summer prior to beginning college), (2) Fall Semester Assessment, (3) Spring Semester Assessment. The pre-college assessment took place in the afternoon and was approximately 90 minutes in length. Fall semester and spring semester assessments were each approximately 60 minutes long.

Pre-College Assessment

Parents accompanied their college-age child to the pre-college assessment during which saliva samples for cortisol assay were taken at four time points (20 minute intervals) from each family member. In order to control for time of day, all pre-college assessments took place in the late afternoon/early evening (4 PM to 7 PM). We evaluated families in the afternoon because afternoon/early evening cortisol varies more in response to transient events than in response to long-term factors (Kirschbaum et al., 1990). We assess cortisol response around public and semi-private family time. The first saliva sample was taken when participants first entered the laboratory, immediately following consenting to study participation. Although each family member worked independently on their own questionnaires, family members sat together in the laboratory waiting room for this task. Initial entry into the laboratory requires one to adapt to a novel environment and self disclose personal information about one’s family and self when completing questionnaires. Participants then completed a series of questionnaires assessing their perceptions of family functioning and individual behavioral adjustment. A second saliva sample was taken from each family member immediately prior to the start of the family interaction task. Family members were escorted into a separate room where they were instructed to build a house of cards together using as many cards as possible. Three decks of cards and a tablecloth were provided in a clear plastic container. Even with the tablecloth, this was a very difficult task that was not amenable to control by participants (Dickerson & Kemeny, 2004). Families were aware of (and had consented to) the video recording of the family interaction task. A third saliva sample was taken at the conclusion of the 20-minute family interaction task. Participants returned to the waiting room and continued work on their questionnaire packets. A fourth saliva sample was taken from each family participant 20 minutes after the conclusion of the family interaction task. Participant families were reimbursed $75 ($25 per family member) for their time engaging in the pre-college assessment.

Fall and Spring Semester Assessments

Student participants returned for a second assessment session during the Fall semester and a third assessment session during the Spring semester of their first college year. Participants were reimbursed $25 for the time they spent engaging in study activities for each follow-up assessment ($50 total). To reduce attrition, we used e-mail and phone reminders about assessment appointments during the semester between follow-up sessions. Follow-up assessment sessions were scheduled for the middle of the Fall and Spring semesters in order to give students time to adjust to the new semester and to avoid assessing participants during the last 3 weeks of the semester when academic stress is often high. During the follow-up assessment sessions, students completed a series of self-report instruments assessing their perceptions of behavioral adjustment.

Measures

Cortisol

To obtain saliva samples we gave each individual a Salivette (Salimetrics, PA), consisting of a large tube that contains a smaller tube with a cotton roll in it. Participants were instructed to remove the cotton roll from the Salivette, hold it in their mouth until it is saturated, then return the cotton roll to the Salivette. We instructed participants to handle only their own samples. Each individual’s sample vial was collected by research assistants. Salivary cortisol immunoassays were performed at a behavioral neuroscience laboratory at the State University of New York, Albany (cohorts 1 and 2) and by Salimetrics LLC (cohort 3). Both laboratories measure free (unbound) salivary cortisol with 7% intra-assay and 15% interassay variance. There were no statistical differences in mean cortisol response or variability among study cohorts.

Family Structure and Process Coding System

The Family Structure and Process (FSAP) coding system (Johnson, Cowan, & Cowan, 1999; Johnson, 2003; Johnson, 2001; Teichman, Cowan, & Cowan, 1988) was used to assess family relatedness during the house of cards task (see Gans & Johnson, in press). The FSAP relies on both triadic-level and dyadic-level assessment of family functioning observed during the family interaction. Trained observers assessed family structure by rating five items: strength of the marital subsystem, differentiation of the marital subsystem, family level cohesion, strength of each parent-emerging adult subsystem. Trained observers also assessed family process by rating six items: positive affect, negative affect, verbal expressiveness, criticism, humor, and praise for each family dyad and at the level of the whole family. Ratings were made using a 5-point scale (1 - very low, 5 = very high). Two raters coded each family independently. Raters then compared scores on each item, reconciling disagreement. Initial interrater reliability (before reconciliation) for each variable ranged from r = .64 - .88.

Eight composite variables were created reflecting the positivity and negativity in each family dyad, and at the level of the triad. High scores on positive dyadic scales (parent-child and mother-father dyads) reflect the presence of strong dyadic subsystems, and dyadic relationships characterized by the expression of positive affect, verbal expressiveness, praise, and humor. High scores on negative dyadic scales (parent-child and mother-father dyads) reflect the presence of negative affect and criticism in the dyadic relationship. High scores on positive triadic family interaction characterize cohesive families and the presence of positive affect, verbal expressiveness, praise, and humor at the level of the triad. High scores on negative triadic family interaction characterize more distant families that express negative affect and criticism.

To identify distinct family types from observations of family functioning, scores on each FSAP scale were cluster analyzed using hierarchical agglomerate clustering method with cosine as the similarity metric determining cluster membership (see Gans & Johnson, in press). Results identified two distinct family types: a cluster of 39 distant and negative families receiving low ratings on the four scales assessing positive family relationships (i.e., positive mother-child dyad, positive father-child dyad, positive mother-father dyad, and positive triad) and high ratings on the four scales assessing negative family relationships (i.e., negative mother-child dyad, negative father-child dyad, negative mother-father dyad, and negative triad); and a cluster of 38 cohesive and positive families receiving high ratings on all four of the positive scales and low ratings on all four of the negative scales.

Family Environment Scale

The Family Environment Scale (FES; Moos & Moos, 2009) measures participants’ perceptions of their family environment by asking participants to respond to statements about their family as true or false. Responses are added to create ten 9-item scales assessing Cohesion, Expressiveness, Conflict, Independence, Achievement Orientation, Intellectual-Cultural Orientation, Active-Recreational Orientation, Moral-Religious Emphasis, Organization, and Control within the family environment. Only the Family Cohesion scale (which measures the commitment, help, and support family members provide for one another), the Family Expressiveness scale (which measures the extent to which family members are encouraged to act openly and to express their feelings directly), and the Family Conflict scale (which measures open conflict in the family) were used in the present study. Raw scores for each scale were converted to standard scores using procedures outlined by Moos & Moos (2009). Scores were then summed to create an overall score of self-reported Family Relatedness.

Internalizing Behavior

The Adult Self-Report Form (ASR, Achenbach & Rescorla, 2003) incorporates many items of the 1997 editions of the Young Adult Self-Report (YASR), plus new items and new national norms that span ages 18-59. Participants rate 123 items that describe people. Participants score the truthfulness of each item in describing their behavior in the past 6 months using a three-point scale (not true, somewhat or sometimes true, and very true or often true). Three additional questions ask participants to indicate the frequency of their tobacco use (In the past 6 months, about how many times per day did you use tobacco?), alcohol use (In the past 6 months, on how many days were you drunk?), and drug use for nonmedical purposes (In the past 6 months, on how many days did you use drugs for nonmedical purposes?). Like the YASR, the profiles for scoring the ASR include normed scales for adaptive functioning, empirically based syndromes, substance use, Internalizing, Externalizing, and Total Problems. The profiles display scale scores in relation to norms for each gender at ages 18-35 and 36-59, based on national probability samples. Student participants completed the ASR during all 3 assessment sessions (e.g., Pre-College, Fall semester, Spring semester). In the present study we focused our analyses on two scales assessing internalizing behavior: depressive problems, and anxiety problems.

Results

Because we sampled cortisol around a challenge task (rather than assessing morning or evening cortisol, or diurnal patterns of cortisol secretion) we characterized each parent participant by his or her cortisol trajectory in response to a family-level challenge. We used latent growth curve modeling (Mplus) of LN transformed data to determine the existence of individual differences in cortisol response to a family interaction task for both mothers and fathers separately (see Luecken, Kraft, & Hagan, 2009 for one example of this approach). We characterized the trajectory of each parent’s cortisol response by testing four growth models: two linear growth models and two nonlinear growth models. To identify our linear model, each slope factor was scaled to reflect constant change between each cortisol level obtained. The first linear model placed the intercept at the first cortisol collection point and the second linear model placed the intercept at the last (fourth) cortisol collection point. To test for the possibility of nonlinear change we tested two models. Once again, the first nonlinear model placed the intercept at the first cortisol collection point and the second nonlinear model tested placed the intercept at the last (fourth) cortisol collection point. Total fit indices were used to determine which model fit the data best (Measelle, Stice, & Hogansen, 2006).

For mothers’ cortisol response to the family interaction, a linear model with the intercept identified as the 4th cortisol level taken achieved the best fit (Chi Square (4df) = 4.94, p = 0.29; CFI = 0.97, TLI = 0.96, RMSEA = 0.49). Mean slope across mothers was −0.02 (SD = 0.08, range = −0.17 to 0.29) and mean intercept was −2.68 (SD = 0.82, range = −5.46 to −0.77). The linear model with the intercept identified as the 1^st cortisol level taken from mothers was not a good fit for these data (chi square = 3.77, p = 0.05; CFI = 0.72, TLI = −0.66; RMSEA = 0.17; SRMR = 0.01). Neither of the non-linear models using mothers’ cortisol data converged.

For fathers’ cortisol response to the family interaction, a quadratic model with the intercept identified as the 1st cortisol level taken achieved the best fit (Chi Square (3df) = 2.04, p = 0.15, CFI = 0.99, TLI = 0.95, RMSEA = 0.08). Mean slope across fathers was 0.12 (SD = 0.85, range = −3.14 to 3.14), mean intercept was −2.44 (SD = 1.14, range = −2.44 to −0.87), mean quadratic was −0.04 (SD = 0.18, range = −0.77 to 0.64). Linear models of fathers’ cortisol resulted in poor fit (Intercept at T1: Chi Square (1df) = 3.08, p = 0.68, CFI = 01.00, TLI = 1.00, RMSEA = 0.11; Intercept at T4: Chi Square(1df) = 71.36, p ≤ 0.001, CFI = 0.94, TLI = 0.02; RMSEA = 0.36). Similarly, the non-linear model of fathers’ cortisol with intercept identified as the last (fourth) cortisol level taken was also a poor fit for these data (Chi Square (3df) = 1.20, p = 0.27, CFI = 1.00, TLI = 0.94, RMSEA = 0.14).

Mean parent cortisol trajectories are marked by a decline in cortisol between the 2^nd and 3^rd cortisol collection points, the points surrounding the family interaction task, with a slight increase during the final 20-minutes of the assessment session. Independent samples t-tests indicate no significant differences in mothers’ and fathers’ mean cortisol slope or intercept for male (mothers’ slope = −0.01, SD = 0.08; mothers’ mean intercept = −2.61, SD = 0.84; fathers’ mean slope = 0.02, SD = 0.71; fathers’ mean intercept = −2.27, SD = 1.15) and female student participants (mothers’ mean slope = −0.02, SD = 0.08; mothers’ mean intercept = −2.71, SD = 0.82; fathers’ mean slope = 0.17, SD = 0.92; fathers’ mean intercept = −2.53, SD = 1.14). Similarly, we found no significant difference in fathers’ mean cortisol quadratic for fathers of sons or daughters (Mean = −0.02, SD = −.14 for sons; Mean = −0.05, SD = 0.14 for daughters).

Emerging Adults’ cortisol response trajectories were also determined using latent variable growth curve modeling (see Gans & Johnson, in press). For student cortisol response to the family interaction, a linear model with the intercept identified as the 4^th (final) cortisol level taken achieved the best fit (chi square = 3.65, p = 0.45; CFI = 1.00, TLI = 1.00; RMSEA = 0.00; SRMR = 0.01). Mean slope across student participants was −0.09 (SD = 0.09, range = −0.31 to 0.19) and mean intercept was −2.36 (SD = 0.70, range = −4.05 to −0.13).

Correlations between mothers’ and fathers’ cortisol response trajectory variables (e.g., intercept, slope, quadratic) indicate a significant correlation between mothers’ and fathers’ cortisol response intercept (r = 0.30, p ≤ 0.01). When fathers had high cortisol levels when beginning our laboratory challenge their wives had a high cortisol level at the end of the family interaction session. Correlations among emerging adults’ cortisol response trajectory variables and the cortisol response trajectory variables of their mothers and fathers indicate significant correlation between emerging adults’ cortisol intercept and the cortisol intercept of their mothers (r = 0.43, p ≤ 0.01) and fathers (r = 0.24, p ≤ 0.05). Emerging adults with higher cortisol levels at the end of the family interaction have mothers who also have higher cortisol levels at the end of family interaction and fathers with higher cortisol values at the start of the interaction. Significant correlations were also indicated between emerging adults’ cortisol response trajectory slope and fathers’ cortisol response trajectory slope (r = −0.22, p ≤ 0.05) and quadratic (r = 0.22, p ≤ 0.05). Emerging adults who showed greater decrease in cortisol levels across the family interaction session had fathers who (a) tended to show significant increases in cortisol response across family interaction but (b) also significant recovery. No other significant correlations were indicated.

To characterize mothers and fathers with respect to their cortisol response slope, intercept, and quadratic, standardized slope, intercept, and quadratic scores were cluster analyzed using hierarchical agglomerate clustering method with cosine as the similarity metric determining cluster membership. Cluster analyses were used to conserve power with our relatively small sample size. Results identified two distinct cortisol response types for both mothers and fathers (see figure 1): a cluster of participants whose cortisol response increased during family interaction (27 mothers; 53 fathers) and a cluster of participants whose cortisol response decreased during family interaction (71 mothers, 48 fathers). Parents with increasing cortisol response during family interaction have a positive slope (Mean = 0.08, SD = 0.07 Mothers; Mean = 0.74, SD = 0.55 Fathers) and low intercept values (Mean = −2.46, SD =0.71 Mothers; Mean = −3.28, SD = 0.76 Fathers). Parents with decreasing cortisol response during family interaction have negative slope (Mean = −0.06, SD = 0.05 Mothers; Mean = −0.44, SD = 0.66 Fathers) and low intercept values (Mean = −2.77, SD = 0.85 Mothers; Mean = −1.68, SD = 0.86 Fathers). Fathers with increasing cortisol response during family interaction also had larger quadratic values (Mean = 0.16, SD = 0.13) compared to fathers with decreasing cortisol response during family interaction (Mean = 0.07, SD = 0.14). Chi-square analysis does not find a significant relationship between parents’ cortisol response to family interaction (chi square = 0.01, p = 0.98), indicating that parents’ cortisol response patterns to family interaction are largely independent of one another.

Figure 1.

Cortisol response type for mothers (top figure) and fathers (bottom figure). There were two overall patterns of salivary cortisol change during the assessment, with a subset of parent participants demonstrating a consistent downward slope between assessments 2 and 4 (- -▼- -), and the remainder demonstrating a consistent flat or increasing slope (–△–).

Parent and Emerging Adult Cortisol Response

We used separate hierarchical multiple regression to test our first hypothesis that parent cortisol response to family interaction will predict student cortisol response to the same family interaction. Separate equations were used to predict student cortisol response intercept and student cortisol response slope to family interaction. Mothers’ and fathers’ cortisol response intercept and cortisol response slope and fathers’ cortisol response quadratic were entered into each equation. For models with significant overall R², we assessed whether parent cortisol response intercept, slope, and/or quadratic make a significant unique contribution to predicting student cortisol intercept and student cortisol slope (i.e., make a significant contribution to predicting student cortisol response over and above the variability predicted by other parent cortisol response variables).

Together, parent cortisol response to family interaction explains approximately 31% of the variance in students’ cortisol intercept response to family interaction (R² = 0.31; p ≤ 0.001). Mothers’ cortisol response intercept makes a significant unique contribution to understanding variability in student cortisol intercept (Beta = 0.41, t = 3.96, p ≤ 0.05). Fathers’ cortisol response intercept trends toward making a significant unique contribution to understanding variability in student cortisol intercept (Beta = 0.59, t = 1.76, p = 0.08). Parents’ cortisol slope (Beta = −0.14, t = −1.36, p = 0.18 for mothers; Beta = 1.54, t = 0.75, p = 0.46 for fathers) and fathers’ cortisol quadratic (Beta = 1.03, t = 0.55, p = 0.58) did not make unique contributions to understanding variability in EA intercept.

Together, parent cortisol response to family interaction explains approximately 11% of the variance in students’ cortisol slope response to family interaction (R² = 0.11; p = 0.06). Fathers’ cortisol response slope (Beta = −4.98, t = −2.13, p ≤ 0.05) and fathers’ cortisol response quadratic (Beta = −4.32, t = −2.05, p ≤ 0.05) make a significant unique contribution to understanding variability in student cortisol slope. Fathers’ cortisol response intercept (Beta = −0.72, t = −1.87, p = 0.07) and mothers’ cortisol intercept (Beta = −0.21, t = −1.78, p = 0.08) trend toward making a significant unique contribution to understanding variability in student cortisol slope. Mothers’ cortisol slope (Beta = 0.13, t = 1.16, p = 0.25) did not make a unique contribution to understanding EA intercept.

Parent Cortisol and Family Functioning

Two separate analyses tested our hypothesis that parents’ cortisol response to family interaction will be linked to family functioning. First, we used a 2 × 2 (observed family type × gender) MANOVA to examine the relationship between observed family functioning and parent cortisol slope, intercept, and quadratic. Results indicate no significant multivariate effect of gender (F(5, 67) = 0.45, p = 0.81), family type F(5, 67 ) = 0.59, p = 0.71), or the interaction of gender by family type (F(5, 67) = 1.48, p = 0.21).

Second, we looked at subjective perception of family functioning and its link with cortisol response patterns. To do this, we used a 2 × 2 × 2 (mothers’ cortisol response type × fathers’ cortisol response type × gender) MANOVA to examine the relationship between parents’ perceptions of Family Relatedness (FES) and cortisol response type. A significant multivariate interaction effect of fathers’ cortisol response and gender was indicated (F(2, 78) = 4.14, p ≤ 0.05; see figure 3) with significant univariate effects for both mothers’ perceptions of family relatedness (F(1, 79) = 7.14, p ≤ 0.01) and fathers’ perceptions of family relatedness (F(1, 79) = 3.82, p ≤ 0.05). Post-hoc analyses find a significant difference in mothers’ reports of family relatedness depending on fathers’ cortisol response pattern but only in families with sons (q(2,87) = 3.52, p ≤ 0.05). In families with sons, mothers report less Family Relatedness when fathers’ cortisol response decreased during triadic family interaction (Mean = 14.00, SD = 4.07, n = 17) than when fathers’ cortisol response increased (Mean = 16.71, SD = 3.53, n = 11). When fathers’ cortisol levels decreased during family interaction, we also find a significant difference in both mothers’ and fathers perceptions of family relatedness among families with sons and families with daughters (q(2,87) = 3.36, p ≤ 0.05 for mothers’ perceptions of family relatedness; q(2,87) = 4.33, p ≤ 0.01 for fathers’ perceptions of family relatedness). In families with daughters, family relatedness was higher than in families with sons (Mothers’ Family Relatedness: Mean = 16.59, SD = 3.25, n = 27 for daughters, Mean = 14.00, SD = 4.07, n = 17 for sons; Fathers’ Family Relatedness: Mean = 16.43, SD = 2.90, n = 27 for daughters, Mean = 13.18, SD = 3.82, n = 17 for sons). No other significant post-hoc differences were found.

Figure 3.

Change in emerging adult’s anxious behavior across the college transition by mothers’ cortisol response type and mothers’ perceptions of family relatedness (FES). The upper figure depicts families where mothers display increased cortisol during family interaction. In these families, if the mother also perceives the family as having low relatedness, EA increases in anxious behavior across the transition year (–❍–). If mothers’ with increasing cortisol perceive their family as having more relatedness, EAs show an initial increase in anxious behavior, followed by a decrease during the spring semester (- - ● - -). In families where mothers display decreased cortisol during family interactions, all EAs showed a pattern of decreasing anxious behavior upon their transition. But this decrease is far more pronounced for families where mothers perceive less family relatedness (–❍–).

Parent Cortisol and Change in Emerging Adult Internalizing Behavior

To test our third hypothesis that parent cortisol response to a family interaction will predict emerging adult offspring’s adjustment to college, we used mixed 3 × 2 × 2 × 2 ANOVA with time as a within groups factor (pre-college, fall semester, spring internalizing behavior), mother cortisol response type (increasing cortisol response pattern, decreasing cortisol response pattern), father cortisol response type (increasing cortisol response pattern, decreasing cortisol response pattern), and observed family type (distant families, cohesive families) as between groups factors. Results indicate a significant 2 way interaction of time and fathers’ cortisol response type for anxiety problems (F(2, 51) = 3.19, p ≤ 0.05) with a significant within-group effect of the same 2 way interaction (time × fathers’ cortisol response; F(1, 52) = 4.87, p ≤ 0.05; see figure 2). Post-hoc analyses find a significant difference in spring semester anxious behavior for EA students depending on their fathers’ cortisol response to family interaction (q(2,87) = 2.87, ≤ 0.05). Emerging adults whose fathers’ cortisol increased during family interaction report more anxiety (Mean = 55.55, SD = 6.97, n = 29) in the Spring of their first college year than those whose fathers’ cortisol decreased during family interaction (Mean = 53.45, SD = 6.60, n = 31). In this analysis, no significant effect of maternal cortisol response pattern or family type was found for emerging adults’ anxious behavior. No significant relationships among parents’ cortisol response, family type, and depressive behavior were indicated.

Figure 2.

Change in emerging adults’ anxious behavior across the college transition by fathers’ cortisol response type. When fathers’ salivary cortisol increases during family interaction, EA anxious behavior increases across the transition year (–△–). When fathers’ salivary cortisol decreased during family interaction, EA anxious behavior decreases across the transition year (- - ▼ - -)

Regression equations were computed to examine the relationship between parents’ perceived family relatedness, parents’ cortisol response type, and change in depressed and anxious behavior during the college transition. Separate equations were used to predict spring semester depressed behavior and spring semester anxious behavior using mothers’ and fathers’ ratings of family relatedness and cortisol response type (i.e., 4 regression equations total). Pre-college behavior, fall semester behavior, family relatedness, and cortisol type (increasing cortisol response pattern vs. decreasing cortisol response pattern) were entered into each equation. To test for interaction effects, the product of cortisol response and family relatedness was also entered into each equation.

Overall regression equations using mothers’ perceptions of family relatedness and cortisol response were statistically significant (R² = 0.59, p < 0.001 for Depressed Behavior; R² = 0.67, p < 0.001 for Anxious Behavior). Family relatedness, cortisol response, and the interaction term (family relatedness × cortisol response type) did not make a significant unique contribution to predicting change in depressed behavior. Mothers’ perceptions of family relatedness (Beta = −0.68, t = −2.22, p ≤ 0.05), mothers’ cortisol response type (Beta = −0.71, t = −2.12, p ≤ 0.05), and the interaction term (family relatedness × cortisol response Beta = 0.97, t = 2.10, p ≤ 0.05), however, each made a significant contribution to explaining change in anxious behavior across the college transition year (see figure 3). Anxious behavior increased across the college transition for emerging adults’ whose mothers’ had an increase in cortisol response during family interaction and who rated their family environment as low in relatedness (FES). In contrast, emerging adults whose mothers’ had an increase in cortisol response during family interaction, but who perceived their family environment to be more highly related (FES) reported a decline in anxious behavior during their first year in college. A similar decline in anxious behavior was reported for emerging adults whose mothers’ had a decrease in cortisol response during family interaction and who characterized their family environment to be low in family relatedness (FES). Last, emerging adults from highly related families (mothers’ FES) in which mothers’ cortisol response decreased during family interaction reported low levels of anxiety across the college transition.

Overall regression equations using fathers’ perceptions of family relatedness and cortisol response were statistically significant (R² = 0.53, p < 0.001 for Depressed Behavior; R² = 0.65, p < 0.001 for Anxious Behavior). Family relatedness, cortisol response, and the interaction term (family relatedness × cortisol response type) did not make a significant unique contribution to predicting change in depressed or anxious behavior.

Discussion

The combined results of this study suggest that measures of salivary cortisol in families provide a unique perspective on family functioning that has predictive value for understanding outcomes during normative life transitions. First, within limits, parents’ cortisol secretion patterns during family interaction predict their emerging adult child’s cortisol secretion pattern. Parents’ cortisol response to family interaction explained approximately 31% of the variance in their emerging adult children’s cortisol intercept, and 11% of the variance in children’s cortisol slope. Mothers’ (significant at p ≤ 0.05) and fathers’ (tendency at p ≤ 0.08) cortisol intercept made unique contributions to their children’s intercept. Fathers’ slopes and quadratics made significant (at p ≤ 0.05) contributions to their children’s slope. This suggests that both mothers and fathers have a unique impact on their children’s cortisol level (intercept), but that fathers have a unique role in influence patterns of cortisol response across a family interaction.

Prior studies have noted mothers’ influence on child cortisol secretion patterns (Stenius et al., 2008), and one has found linkages among mothers, fathers, and adolescent children in cortisol responding during a family conflict conversation, so that mothers predicted fathers’ cortisol, fathers predicted adolescents’ cortisol, and adolescents predicted mother’s cortisol (Saxbe et al., 2014). The current findings contrast with Saxbe et al. (2014), in that fathers’ cortisol patterns predicted mothers’ cortisol. Specifically, when fathers had high cortisol levels when beginning the laboratory challenge, their wives had high cortisol levels at the end of the family interaction. The divergence of the current results from Saxbe et al. (2014) may be attributable to differences in participants and procedures. Specifically, participants in the Saxbe et al. (2014) study were between 9 and 13 years of age, a range when familial integration and physiological co-influence might be more intense than for families with EA children. In addition, Saxbe et al. (2014) had participants in the laboratory for a longer duration, relaxed participants to obtain a baseline cortisol level, then sampled cortisol intensively around a conflict task. This contrasts markedly with the current procedure, which consisted of a shorter duration laboratory visit, and cortisol sampled around several tasks, none of which included a guided conflict discussion. The commonality between Saxbe et al. (2014) and the current results is that parent response patterns are not independent, and therefore cannot be said to independently affect child response patterns, though each makes their own unique contributions. It would be impossible to see this without the triadic approach to understanding family interaction utilized in this and the Saxbe et al. (2014) study.

Parent cortisol response was not linked with observed family functioning, but was linked with parents’ subjective perceptions of family functioning. This link was moderated by children’s gender. In families with emerging adult women, mothers’ and fathers’ perception of family relatedness was greatest when fathers’ cortisol decreased during the family interaction. The pattern was reversed in families with emerging adult men, so that mothers’ perception of family relatedness was greatest when fathers showed an increase in cortisol during the family interaction. Other work from our laboratory has indicated that emerging adult cortisol response patterns are linked with observed family functioning as well as their own subjective reports of family functioning (Gans & Johnson, in press). Specifically, when there are men (sons or fathers) in the family increased cortisol response predicts greater family relatedness and cohesion. In contrast, with emerging adult women in the family, daughters’ and fathers’ cortisol decreases are linked with greater family relatedness (Gans & Johnson, in press).

These results highlight the importance of incorporating insider and outsider perceptions of family functioning. Each measure has its own predictive value for emerging adults in the current study, and only subjective measures have predictive value for parents. Because there was no way to know this a priori, it was important that both types of variables were incorporated into the present study. These results indicate that parents have two routes through which they influence their children’s psychophysiological responses to challenge; first, through long-established family context, and, second, through their own psychophysiological response within a given setting, which is largely based on their own perception of family functioning, rather than on more objective measures.

Family cortisol patterns have an important role in emerging adult outcomes during the transition to college. When fathers’ cortisol increased during family interaction, their emerging adult child’s self-reported anxious behavior increased across the transition to college. Anxious behavior decreased across the first year of college for emerging adults whose fathers’ cortisol decreased during family interaction. Fathers’ cortisol and perception of family relatedness made significant overall contributions to emerging adult depressed and anxious behaviors, so that emerging adults report more depressed and anxious behaviors when they have fathers with increasing cortisol during the family interaction who report lower family relatedness.

The pattern of results for mothers is more complex, but also more revealing. When mothers’ cortisol increased during the family interaction, and mothers rated their family as low in relatedness, anxious behavior increased across the college transition for emerging adults. In contrast, emerging adults whose mothers’ had an increase in cortisol response during family interaction, but who perceived their family environment to be more highly related reported a decline in anxious behavior during their first year in college. Emerging adults from highly related families in which mothers’ cortisol response decreased during family interaction reported low levels of anxiety across the college transition. This pattern of results suggests that the combination of increasing maternal cortisol during family interaction coupled with self-reported distant family results in an increase in anxious behavior across the first college year. Interestingly, emerging adult cortisol and observed family cohesion interact in the same way, so that emerging adults with increasing cortisol during family interaction and distant families demonstrate a similar increase in anxious behavior across the first college year (Gans & Johnson, in press).

Observed and self-reported family relatedness has often been linked with increased emotional competence (Johnson, 2003; Johnson, Cowan, & Cowan, 1999; Sturge-Apple, Davies, & Cummings, 2010). Similarly, self-report of various aspects of family functioning has been related to cortisol response patterns during challenge (Hibel et al., 2011; Luecken, 1998; Luecken, 2000; Luecken, Kraft, & Hagan, 2009; Rifkin-Graboi, 2008). The novelty of this study is to combine all of these approaches, to begin to elucidate complex interactions among insider and outsider perspectives of family functioning, cortisol response patterns, and emerging adult child outcomes across a relevant developmental transition. These results represent an initial step toward understanding whole family biopsychology obtained in a relatively small, homogenous sample. For example, the relatively small sample studied did not allow for more robust growth mixture models that test differential growth among different classes and cell sizes for interactions were relatively small at times. Instead, we had to rely on cluster analyses to identify different cortisol response patterns among our participants, an analytical technique that does not allow for tests of model fit.

To date, there is little empirical understanding of the conditions that may encourage physiological linkage among family members, the implications of physiological linkage for psychosocial development, or the nature of the relationship among linked physiological systems (e.g., is physiological reactivity contagious or do we compensate for another’s arousal?) (see Timmons, Margolin, & Saxbe, 2015). The relationship found in the present study between mothers’ and fathers’ cortisol response to family interaction offers some tentative data to support the concept of physiological contagion. When fathers cortisol was elevated at the start of the interaction, mothers’ cortisol levels were also elevated by the end of the interaction. A similar pattern of contagion was noted between fathers and their emerging adult offspring. The pattern between mothers and emerging adults appeared more synchronous rather than contagious, with both mothers and their EA children ending the interaction with concordant cortisol levels.

Moreover, these data also suggest that the family context may be an important factor in understanding the nature of physiological synchrony and in understanding the implications of physiological synchrony for psychosocial development. Although concordance in cortisol response patterns was not the focus of these analyses, the findings presented suggest that the relationship among cortisol response and college adjustment (particularly anxiety during the college transition) depend on a combination of factors including perceived family relatedness, observed family functioning, and parents’ physiological response during family interaction. Few studies have examined the role of emotional context in studying the implications of physiological synchrony, and those that have rarely find a significant relationship (see Timmons, Margolin, and Saxbe, 2015). Our findings, however, argue that the emotional context is an important factor in unpacking the relationship between parents’ physiological response to family interaction and emerging adults’ social-emotional adjustment to college.

The results of this study emphasize the importance of this complex approach to the study of family contributions to child outcomes. Our approach allowed us to understand not just how individual physiology contributes to success and struggle during a normative life transition (Gans & Johnson, in press), but also how parent physiology makes a unique contribution to these outcomes. Moreover, insider and outsider perspectives of family functioning provided further moderating influences on outcomes, demonstrating that neither physiology nor family context stand alone as determinants of success during life transitions. Finally, the triadic approach to studying families allowed us to see that fathers make an important contribution to their emerging adult children’s success during their first college year. Though it complicates data collection and interpretation of results, this interdisciplinary, multifaceted approach to the understanding of triadic family dynamics during development provides nuanced results that complicate our understanding of the ways in which individuals develop emotionally and socially. This type of approach provides us with possible biomarkers for success and failure during life transitions, and suggests points of intervention that may help individuals succeed during critical life transitions.