Momentary Work Worries, Marital Disclosure and Salivary Cortisol Among Parents of Young Children

Abstract

Objective

To investigate whether worries about work are linked to people’s own cortisol levels and their spouses’ cortisol levels in everyday life, and whether marital factors may moderate these links. While research has shown that satisfying marriages can buffer the physiological effects of everyday stress, the specific mechanisms through which marriage influences the processing and transmission of stress have not yet been identified.

Methods

Thirty-seven healthy married couples completed baseline measures and then provided saliva samples and indicated their worries about work for 6 times a day from a Saturday morning through a Monday evening.

Results

Wives’ cortisol levels were positively associated with their own work worries (p = .008) and with their husbands’ work worries (p = .006). Husbands’ cortisol levels were positively associated only with their own work worries (p = .015). Wives low in both marital satisfaction and disclosure showed a stronger association between work worries and cortisol compared to wives reporting either high marital satisfaction and/or high marital disclosure.

Conclusions

These results suggest that momentary feelings of stress affect not only one’s own cortisol levels, but affect close others’ cortisol levels as well. Further, they suggest that for women, the stress-buffering effects of a happy marriage may be partially explained by the extent to which they disclose their thoughts and feelings with their spouses.

INTRODUCTION

The word “family” often conjures up images of a stable, restorative place where people can come home to recuperate from life’s daily worries. In reality, the family is a dynamic system, existing not in a vacuum but rather shaped by outside experiences, including stressful ones. It is now well known that reactions to daily stressors can persist after an event has occurred, with individuals carrying home the residue of stressful experiences from work and school into the home—affecting not only peoples’ own mood but the mood of other family members as well (1). For example, research has shown that negative mood at work predicts day-to-day changes in angry marital behavior (2), and more stressful experiences at work predict greater distraction and less responsiveness to spouses at home (3).

While it is clear that experiences at work affect family dynamics at home, the physiological consequences of work-family spillover are less well known. Preliminary findings suggest that stressful experiences at work can affect people physiologically at home, and that satisfying marriages can buffer these effects (4). However, the specific mechanisms through which marriage influences the processing and transmission of stress on a daily basis have not yet been identified. With this article, we examine the links between short-term changes in a common daily stressor—worry about work—and salivary cortisol among working parents. We also addressed how between-couple variation in an important marital process—self-disclosure—may moderate the links between work worries and cortisol.

A large number of human and animal studies have documented the effects of stress exposure on the hypothalamic-pituitary-adrenal (HPA) axis. The HPA axis influences other peripheral physiological systems and ultimately health via cortisol, its principal hormonal product (5, 6). Over the past two decades there has been an increased focus on the effects of stress and mood on cortisol in everyday life, with an estimated 38 naturalistic cortisol studies having been conducted (7). Although these studies have demonstrated links between stress and cortisol in naturalistic settings (8–10), remarkably few studies have investigated how daily stress might affect close others’ cortisol levels. One recent study showed that greater work hours are positively associated with one’s own and one’s spouse’s cortisol (11). However, no studies to our knowledge have examined how fluctuations in concerns about work that people carry home with them (which we term work worries) affect cortisol.

The strength of within-day associations between work worries and cortisol at home may depend in part on the quality of the marital relationship. For example, among married working mothers, evening cortisol declines most sharply for those highest in marital satisfaction (4). Further, higher daily levels of intimacy are linked to lower same-day cortisol levels among dual-earner couples (12). However, very little is known about the specific elements of marital quality that may moderate the effects of stress on the HPA axis. One possible candidate is marital disclosure—the extent to which people open up to their spouses about their thoughts and feelings.

There has been growing interest in self-disclosure in daily life, particularly in the context of close relationships (13–15). For example, higher daily levels of self-disclosure in romantic relationships are associated with increased intimacy in those relationships (14, 15). Surprisingly, the effects of self-disclosure in relationships on stress physiology have largely been overlooked. Actively holding back thoughts, emotions, or behaviors can exacerbate a number of adverse biological processes, such as increased cortisol production and immune suppression (16). A large body of experimental evidence has shown that disclosing one’s thoughts and feelings through expressive writing results in improvements in physical health (17, 18) and HPA axis functioning (19, 20). This evidence suggests that disclosures in everyday life—such as marital disclosure—may also have health benefits. Based on previous findings, we expected marital disclosure to buffer the effects of work worries on cortisol.

The major aim of this work was to investigate how work worries when people are at home with their families are linked to one’s own cortisol and spouses’ cortisol. To our knowledge, no published studies have examined the effects of one person’s psychological state on another person’s cortisol (e.g., one’s spouse) in everyday life. The second aim was to investigate whether marital disclosure moderates the links between momentary work worries and cortisol. While previous studies have reported effects of written emotional disclosure on biological processes, none to our knowledge has examined the effects of disclosure to close others on biology.

In this study, couples completed Ecological Momentary Assessment (EMA)(21) questionnaires assessing work worries and provided saliva samples six times per day from a Saturday morning through a Monday evening, which captured the days of the week (weekends) that families typically spend the bulk of their time together as well as the often stressful transition from leisure to work (Mondays) (22). We conducted dyadic analyses (23) to assess the effects of one’s own feelings of work worries and the effects of spouses’ work worries on cortisol.

Based on previous findings, we hypothesized:

• (H1)Momentary work worries would be positively associated with one’s own cortisol and spouses’ cortisol over the three days of the study.
• (H2)Higher overall levels of marital disclosure would moderate the association between momentary work worries and cortisol. In other words, we expected the effects of momentary work worries on cortisol to be attenuated among individuals who were more self-disclosing with their spouses.

Method

Participants

Fifty-one families from Austin, TX were recruited for a larger study of everyday family life. All families included a mother and a father who were married or cohabitating for at least two years, and had at least one child between the ages of 3 and 5. Analyses were restricted to those couples who indicated that they both worked at least part-time outside the home. Because of known effects of pregnancy on the HPA axis (24), five couples in which the wife was pregnant were excluded from analyses. Based on these criteria, thirty-seven complete couples were included in the final sample. Data was collected for this study from August 2006 through May 2007 under approval from the University of Texas at Austin Institutional Review Board.

The average age of participants was 35.6 for husbands (SD = 4.6) and 34.5 for wives (SD = 4.1). The average length of marriage in the study was 7.4 years (SD = 3.9). Wives worked 30.0 hours per week on average (SD = 14.4), while husbands worked 41.4 hours on average (SD = 17.3). The ethnic make-up of the sample was 73.0% White/Caucasian, 17.6% Hispanic/Latino, 5.4% Black/African American, and 4.0% indicating other ethnicity. The median annual family income was $87,000, ranging from $20,000 to $500,000. Each family was paid $250 for participating in the study.

Procedure

Families initially came to the lab on a weekday evening to complete baseline questionnaires and receive an overview of the study. During the following Saturday morning through Monday evening participants completed momentary questionnaire assessments and provided saliva samples 6 times per day.

Baseline Measures

Marital disclosure

Marital disclosure was assessed with a two-item scale developed for this study in which participants were asked, on a 7-point Likert-type scale (1= very little, 7 = a great deal), “To what extent do you disclose your deepest thoughts and feelings to your spouse?” and, “To what extent do you say what’s on your mind to your spouse.” Alpha reliability for the current sample was .77; mean scores were 5.68 for wives and 4.51 for husbands.

Marital satisfaction

The Marital Adjustment Test (MAT) is a well-validated sixteen-item measure of marital satisfaction (25). Alpha reliability in this study was .79 for wives and .75 for husbands. In this sample, the mean score for wives was 104 (range 47–145) and 105 for husbands (range 40–140). This measure was included as a covariate to determine the unique moderating effects of marital disclosure on the links between momentary work worries and cortisol—above and beyond marital satisfaction—and marital disclosure X marital satisfaction X work worries interaction effects. In this sample, correlations between the MAT and marital disclosure were .46 for husbands (p = .004) and .26 for wives (p = .12).

Health-related variables

To control for possible confounds in the cortisol data, we asked participants to provide information about nicotine habits, caffeine intake, body mass index, and birth control medication.

Ecological Momentary Assessment Measures

Cortisol Collection

From the Saturday morning through Monday evening that families participated in the study, spouses were instructed to self-collect saliva samples and report collection times at six time points: immediately upon wakening, 45 minutes later (prior to any eating, drinking, or exercise), at 3 semi-random beeped time points in the early evening (approximately 5 pm, 7 pm and 9 pm), and then at bedtime. The timing of these samples corresponds to recommendations by the MacArthur Research Network on Socioeconomic Status and Health (26). Spouses were each given a signaling device (Casio DataBank DBC-60 programmable watch, Casio USA, Dover, NJ), which was preprogrammed to emit signals during the 3 semi-random times each day. Saliva was collected using a Salivette (Sarstedt 1534, Sarstedt Inc., Newton, NC), consisting of a sterilized cotton swab that the participant placed in his/her mouth for 2 minutes and then stored in a small beaker contained in a plastic tube.

Substantial efforts were made to emphasize the importance of compliance with the study’s procedures, particularly with regard to the timing of saliva sampling immediately upon waking (27). These efforts included having participants take a practice sample at the outset of the study, explaining to participants why exact timing of the samples is essential, and asking participants to note any sampling issues that occurred. Participants were asked to refrigerate Salivette samples at the end of each day and placed them in plastic bags provided by the experimenter. Upon returning the Salivettes to the lab the following week, samples were stored in a −20° C freezer.

Cortisol levels were determined by time-resolved immunoassay with fluorometric end point detection (28) at the Biological Psychology laboratory directed by Dr. Clemens Kirschbaum at the Technical University of Dresden in Dresden, Germany. The test used 50 ul of saliva and had average intra- and inter-assay coefficients of variation of less than 8%. To correct for positive skewness a natural log₁₀ transformation was performed on the cortisol data prior to analyses. A constant of 1 was added to cortisol values prior to transformation so that all transformed values would be positive.

Momentary work worries

At the six points each day that they provided saliva samples, participants completed an EMA measure of work worries adapted from previous research (29, 30). At each sampled moment, participants were given three paired descriptors indicating how they felt about their job at that moment: “Worried about/Not worried at all” (reverse-scored), “Relaxed about/Tense toward”, and “In Control of/Controlled by.” The descriptors were rated on a 7-point Likert-type scale (ranging from −3 to +3), standardized to a mean of 0 and a SD of 1, and then averaged together to form a work worries composite. Alpha reliability for the scale in the current sample was .85, with mean scores of 0.08 (SD = .63) for wives and −0.09 (SD = .59) for husbands.

Momentary negative affect

To rule out the possibility that any effects of momentary work worries on cortisol were simply due to overall negative affect (NA), a well-validated 13-item measure of momentary NA (29) was completed at each time point. In response to the question, “How are you feeling right now?” the respondent rated words such as “frustrated,” “worried,” and “irritable” on a 4-point Likert-type scale (0 = not at all, 3 = very much), standardized to a mean of 0 and a SD of 1, and then averaged together to form a composite measure of NA. Alpha reliability for this scale in the current sample was .87, with mean scores of 0.00 for wives (SD = .38) and −0.04 for husbands (SD = .36).

Daily Perceived Stress

To rule out the possibility that any effects of momentary work worries were due to higher overall levels of stress that day, participants completed a short form of the Perceived Stress Scale (PSS) (31) each night. The short form PSS consists of 4 items that ask respondents how they felt over the past 24 hours (e.g., “In the past 24 hours, how often have you felt that you were unable to control the important things in your life?”); responses were provided on a five-point Likert-type scale (0 = never, 4 = fairly often). Alpha reliability for the current sample for this measure was .74, with mean scores of 1.30 for wives (SD = .70) and 1.25 for husbands (SD = .61).

Overview of Data Analytic Strategy – The Actor-Partner Interdependence Model (APIM)

A unique characteristic of marital data is that the data from spouses are not independent. For example, people who are satisfied in their marriage tend to have spouses who also are satisfied; people who are optimistic tend to have optimistic spouses, and so on. To account for this data interdependence in statistical analyses, researchers in recent years have begun to frame their analyses in the Actor-Partner Interdependence Model (APIM)(32). The APIM is a technique designed to address interdependence in dyadic (couple) data. This technique allows researchers to estimate, for example, the influence of one person’s momentary work worries on her own cortisol—called actor effects—as well as the effects of her momentary work worries on her spouse’s cortisol—called partner effects. We illustrate this basic APIM design in Figure 1.

Figure 1.

Actor-Partner Interdependence Model (APIM) used estimate the association between momentary experiences and cortisol.

APIM analyses in this study were conducted using Hierarchical Linear Modeling (33). Because of the strong diurnal rhythm of cortisol, HLM is ideal for its analysis (34). Further, HLM can estimate slopes and intercepts even with missing data¹. All APIM analyses were conducted using the longitudinal dyadic analysis model proposed by Laurenceau and Bolger (35). With this type of model, data from husbands and wives are on separate lines in the data files, nested within couple-level IDs. Separate intercept and slope terms are created for husbands and wives, with spouses denoted by dummy variables used to calculate intercepts. Using this type of statistical technique, one can model, for example, the effects of wives’ work worries on their own cortisol levels and on their husbands’ cortisol levels and the effects of husbands’ work worries on their own cortisol levels and on their wives’ cortisol levels, simultaneously estimating all four of these effects.

A three-level multilevel growth-curve analysis (36) was used. Variables that change with each cortisol sample, such as momentary work worries, are Level 1 variables; day-to-day experiences, such as daily perceived stress and weekends vs. weekdays, are Level 2 variables; person-level variables, such as marital disclosure and marital satisfaction, are Level 3 variables. We included both same-sampling moment and prior-sampling moment NA and work worries as Level 1 predictors. This allowed us to examine whether cortisol at each moment was more strongly associated with one’s own experiences and spouses’ experiences at the prior sampling moment vs. sampling moment concurrent with cortisol sampling. Effect sizes were computed from t values (effect size r = sqrt(t²/t²+df)) and are reported in table in the results (37).

Because of the large lag between the second and third sampling occasions (from early morning to early evening), the second sampling occasion was not included as a prior-moment predictor in analyses. However, we were still able to incorporate the second-sampling occasion work-worries as a same-moment predictor of cortisol (just not as a prior-moment predictor). Additionally, only same-moment experience data were available for the first sampling occasion of each day. Prior-moment work worries and negative affect were assessed, on average, 1.37 hours earlier, with 93% of the assessments no more than 2 hours earlier (SD = 1.01).

First, a Level 1 (moment-level) model was fit to provide estimates of the parameters defining each person’s diurnal cortisol rhythm:

$$\text{Level}1:\text{Cortisol}={\text{Husband}(\mathrm{\pi }1}_{\mathrm{i}}\text{Intercept}+{\mathrm{\pi }3}_{\mathrm{i}}{\text{Time}\text{Since}\text{Waking}}_{\text{ij}}+{\mathrm{\pi }5}_{\mathrm{i}}{\text{Time}\text{Since}\text{Waking}}_{\text{ij}}^2+\mathrm{\pi }{7}_{\mathrm{i}}{\text{Cortisol}\text{Awakening}\text{Response}}_{\text{ij}})+\text{Wife}(\mathrm{\pi }{2}_{\mathrm{i}}\text{Intercept}+{\mathrm{\pi }4}_{\mathrm{i}}\text{Time}{\text{Since}\text{Waking}}_{\text{ij}}+{\mathrm{\pi }6}_{\mathrm{i}}{\text{Time}\text{Since}\text{Waking}}_{\text{ij}}^2+\mathrm{\pi }{8}_{\mathrm{i}}{\text{Cortisol}\text{Awakening}\text{Response}}_{\text{ij}})+{\mathrm{\epsilon }}_{\text{ij}}$$ (Equation 1)

Salivary cortisol shows a strong basal diurnal rhythm, with levels typically high in the morning upon wakening, sharply increasing in the first 30–45 minutes after awakening, dropping rapidly over the next few hours, and then declining more slowly across the day. We thus modeled cortisol values as a function of the time of each sample, scaled as hours since time waking each day, such that Level 1 intercepts reflected an estimate of each person’s average wakeup cortisol level across the days of testing. Both linear and quadratic terms for time of day (“Time Since Waking” and “Time Since Waking²”, respectively, in the above equation) were included to account for curvilinear nature of diurnal cortisol slopes. In line with previous research (10), the 45-min after awakening cortisol sample was indicated with a dummy variable (0, 1), with the coefficient on that variable (π7_i for husbands and π8_i for wives) representing an estimate of participants’ cortisol awakening response (CAR). Previous research indicates that a higher CAR is associated with greater psychological stress (38) and perceived work overload (39), and is greater on weekdays compared to weekends (39, 40).

Second, we estimated Level 2 (day-level) effects of daily perceived stress and the effects of weekend days vs. Mondays on cortisol:

$$\text{Level}2:\mathrm{\pi }1\text{to}\mathrm{\pi }8={\mathrm{\beta }}_{\mathrm{i}0}\text{to}{\mathrm{\beta }}_{\text{ij}}\mathrm{x}\text{Day-Level Variables}+r_{\text{ij}}$$ (Equation 2)

Third, we estimated Level 3 (day-level) effects of person-level variables known to influence cortisol. These variables included health-related variables (nicotine and caffeine intake, BMI, age, and birth control medication) and other relevant person-level variables (average number of hours worked each week, annual income, and time married):

$$\text{Level}3:{\mathrm{\beta }}_{\mathrm{i}0}\text{to}{\mathrm{\beta }}_{\text{ij}}={\mathrm{\gamma }}_{\text{ij}0}+{\mathrm{\gamma }}_{\text{ijk}}\mathrm{x}\text{Person-Level Controls}+{\mathrm{\mu }}_{\text{ij}}$$ (Equation 3)

Fourth, momentary predictors (NA and work worries) were entered at Level 1, simultaneously estimating actor and partner effects of both prior sampling-moment and same sampling-moment experiences (e.g, NA and work worries at each moment). The example of momentary NA is used here:

$$\text{Level}1:\text{Cortisol}=\mathrm{\pi }1+\dots \mathrm{\pi }8+\text{Husband}(\mathrm{\pi }{9}_{\mathrm{i}}{\text{Husband Same-Moment NA}}_{\text{ij}}+\mathrm{\pi }{10}_{\mathrm{i}}{\text{Husband Prior-Moment NA}}_{\text{ij}}+\mathrm{\pi }{11}_{\mathrm{i}}{\text{Wife Same-Moment NA}}_{\text{ij}}+\mathrm{\pi }{12}_{\mathrm{i}}\text{Wife Prior-Moment}{\text{NA}}_{\text{ij}})+\text{Wife}(\mathrm{\pi }{13}_{\mathrm{i}}{\text{Wife}\text{Same-Moment}\text{NA}}_{\text{ij}}+\mathrm{\pi }{14}_{\mathrm{i}}{\text{Wife}\text{Prior-Moment}\text{NA}}_{\text{ij}}+\mathrm{\pi }{15}_{\mathrm{i}}\text{Husband}{\text{Same-Moment}\text{NA}}_{\text{ij}}+\mathrm{\pi }{16}_{\mathrm{i}}{\text{Husband}\text{Prior-Moment}\text{NA}}_{\text{ij}})+{\mathrm{\epsilon }}_{\text{ij}}$$ (Equation 4)

Finally, the moderating effects of the marital variables (marital disclosure and marital satisfaction) were included at Level 3:

$$\text{Level}3:{\mathrm{\beta }}_{\mathrm{i}0}\text{to}{\mathrm{\beta }}_{\text{ij}}={\mathrm{\gamma }}_{\text{ij}0}+{\mathrm{\gamma }}_{\text{ijk}}\mathrm{x}\text{Marital}\text{Variables}{\mathrm{+\mu }}_{\text{ij}}$$ (Equation 5)

Results

Initial Models of Diurnal Cortisol with Effects of Time, Time², CAR and Day-Level Predictors

We first ran an HLM model with time, time² and CAR as predictors (Equation 1 above); all effects were significant and are displayed in Table 1. As illustrated in Figure 2, cortisol levels showed a decline across the day (effect of time) and a slight end-of-day increase (effect of time²). Next, we entered Level 2 (day-level) predictors into the model (a combination of Equations 1 and 2); results are displayed in Table 2. Based on criteria from previous EMA cortisol research (8), variables with effects of p < .20 were included as covariates in all subsequent analyses. Using this criteria, higher wives’ daily perceived stress (PSS scale) was associated with husbands’ flatter cortisol slopes (effect of time) and steeper cortisol increases at the end of the day (effect of time²). Husbands’ perceived stress was associated with a smaller CAR for wives. The only significant day-level effects were the effects of weekends vs. Mondays, with both husbands and wives showing greater CARs on Mondays compared to weekends.

Table 1.

Initial Model of Diurnal Cortisol

Fixed effect	Coefficient (Standard Error)	T ratio	p
Husbands’ intercept, π1	1.024 (0.031)	33.21	.001
Wives’ intercept, π2	0.963 (0.036)	26.53	.001
Husbands’ average slope of time since waking, π3	−0.076 (0.006)	−12.84	.001
Wives’ average slope of time since waking, π4	−0.070 (0.009)	−8.04	.001
Husbands’ average slope of time since waking2, π5	0.002 (0.000)	4.90	.001
Wives’ average slope of time since waking2, π6	0.002 (0.000)	3.27	.003
Husbands' average CAR, π7	0.067 (0.026)	2.27	.01
Wives' average CAR, π8	0.037 (0.022)	2.97	.006

Note. Intercepts indicate average cortisol values at wakeup; average slopes of time since waking indicate change in cortisol per 1-hour change in time; average slopes of time since waking² indicate change in cortisol per 1-hour change in time²; CAR = Cortisol Awakening Response, indicating amount of change in cortisol during the 45 minutes after waking.

Figure 2.

Cortisol values for all participants over the three days of the study.

Table 2.

Effects of Day-Level Predictors

Fixed effect	Coefficient (Standard Error)	T ratio	Effect Size r	p
Husbands’ intercept, β10	1.041 (0.031)	33.51		.001
Weekends vs. Mondays, β11	−0.047 (0.060)	−0.79	.13	.44
Wives' daily perceived stress, β12	−0.046 (0.042)	−1.08	.18	.83
Husbands' daily perceived stress, β13	−0.016 (0.057)	−0.28	.05	.78
Wives’ intercept, β20	0.950 (0.043)	21.89		.001
Weekends vs. Mondays, β21	0.042 (0.046)	0.92	.15	.36
Wives' daily perceived stress, β22	−0.031 (0.039)	−0.78	.13	.43
Husbands' daily perceived stress, β23	−0.014 (0.053)	−0.26	.04	.79
Husbands’ average slope of time since waking, β30	−0.076 (0.006)	−12.87		.001
Weekends vs. Mondays, β31	0.000 (0.013)	0.03	.00	.98
Wives' daily perceived stress, β32	0.026 (0.014)	1.82	.29	.08
Husbands’ daily perceived stress, β33	0.012 (0.013)	0.91	.15	.37
Wives’ average slope of time since waking, β40	−0.066 (0.009)	−6.95		.001
Weekends vs. Mondays, β41	−0.015 (0.012)	−1.20	.20	.24
Wives' daily perceived stress, β42	0.015 (0.013)	1.10	.18	.28
Husbands' daily perceived stress, β43	−0.011 (0.019)	−0.58	.10	.56
Husbands’ average slope of time since waking2 β50	0.001 (0.000)	4.05		.001
Weekends vs. Mondays, β51	0.001 (0.001)	0.94	.15	.35
Wives' daily perceived stress, β52	−0.002 (0.001)	−1.91	.30	.06
Husbands' daily perceived stress, β53	−0.001 (0.001)	−0.92	.15	.36
Wives’ average slope of time since waking2, β60	0.001 (0.001)	2.42		.02
Weekends vs. Mondays, β61	0.001 (0.001)	1.46	.24	.15
Wives' daily perceived stress, β62	−0.001 (0.001)	−0.85	.14	.40
Husbands' daily perceived stress, β63	0.001 (0.001)	0.60	.10	.55
Husbands' average CAR, β70	0.011 (0.029)	0.37		.71
Weekends vs. Mondays, β71	0.189 (0.071)	2.66	.40	.01
Wives' daily perceived stress, β72	0.038 (0.050)	0.76	.13	.45
Husbands' daily perceived stress, β73	−0.018 (0.082)	−0.22	.04	.83
Wives' average CAR, β80	0.040 (0.028)	1.40		.16
Weekends vs. Mondays, β81	0.090 (0.040)	2.28	.36	.03
Wives' daily perceived stress, β82	0.051 (0.040)	1.25	.21	.21
Husbands' daily perceived stress, β83	−0.136 (0.068)	−1.99	.32	.06

Note. Intercepts indicate average cortisol values at wakeup; average slopes of time since waking indicate change in cortisol per 1-hour change in time; average slopes of time since waking² indicate change in cortisol per 1-hour change in time²; CAR = Cortisol Awakening Response, indicating amount of change in cortisol during the 45 minutes after waking. Bolded rows indicate covariates that were included in subsequent analyses (p < .20).

Analyses to Determine Inclusion of Person-Level Covariates

We next added Level 3 (person-level) health covariates and work, marital and financial covariates to the model (a combination of Equations 1–3). The results of these analyses are shown in Tables 3 and 4, respectively. Notable effects (p < .20) in these analyses included: 1) greater wives’ hours worked per week was related to lower wives’ wake-up cortisol, a less steep end-of-the-day cortisol increase, and lower husband’s wake-up cortisol, 2) greater wives’ hours per week was related to flatter wives’ cortisol slope, and 3) greater length of marriage was related to higher wives’ wake-up cortisol. Together, these findings indicate that wives’ greater work hours are associated with less “healthy” diurnal cortisol profiles for both husbands and wives, and, to a lesser extent, that length of marriage appears to be associated with a healthier cortisol profile for wives.

Table 3.

Effects of Person-Level Health Predictors

Fixed effect	Coefficient (Standard Error)	T ratio	Effect Size r	p
Husbands’ intercept, γ100	1.0227 (0.0305)	33.49		.001
BMI, γ101	0.0010 (0.0037)	0.29	.05	.78
Tobacco/nicotine, γ102	−0.0150 (0.0100)	−1.51	.25	.14
Caffeine, γ103	−0.0073 (0.0270)	−0.27	.12	.79
Age, γ104	−0.0043 (0.0065)	−0.66	.11	.51
Wives’ intercept, γ200	0.9727 (0.0376)	25.82		.001
BMI, γ201	0.0038 (0.0068)	0.56	.09	.58
Birth control, γ202	−0.0826 (0.1110)	−0.77	.13	.46
Caffeine, γ203	−0.0192 (0.0423)	−0.46	.08	.65
Age, γ204	0.0020 (0.0089)	0.23	.04	.82
Husbands’ average slope of time since waking, γ300	−0.0745 (0.0061)	−12.03		.001
BMI, γ301	−0.0004 (0.0007)	−0.54	.09	.59
Tobacco/nicotine, γ302	0.0051 (0.0020)	2.50	.39	.02
Caffeine, γ303	−0.0056 (0.0057)	−0.99	.16	.33
Age, γ304	−0.0010 (0.0013)	−0.77	.13	.45
Wives’ average slope of time since waking, γ400	−0.0718 (0.0086)	−8.29		.001
BMI, γ401	−0.0037 (0.0016)	−2.29	.36	.03
Birth control, γ402	0.0120 (0.0270)	0.45	.07	.66
Caffeine, γ403	0.0144 (0.0098)	1.470	.24	.15
Age, γ404	−0.0013 (0.0021)	−0.66	.11	.51
Husbands’ average slope of time since waking2, γ500	0.0016 (0.0004)	4.30		.001
BMI, γ501	0.0000 (0.0000)	0.72	.12	.48
Tobacco/nicotine, γ502	−0.0002 (0.0001)	−1.81	.29	.08
Caffeine, γ503	0.0003 (0.0004)	0.89	.15	.39
Age, γ504	−0.0000 (0.0000)	−0.51	.08	.61
Wives’ average slope of time since waking2, γ600	0.0016 (0.0005)	3.29		.003
BMI, γ601	0.0002 (0.0000)	2.51	.39	.02
Birth control, γ602	0.0000 (0.0015)	0.02	.00	.99
Caffeine, γ603	−0.0009 (0.0005)	−1.63	.26	.11
Age, γ604	0.0001 (0.0001)	0.80	.13	.43
Husbands' average Cortisol Awakening Response, γ700	0.0650 (0.0282)	2.31		.03
BMI, γ701	−0.0020 (0.0035)	−0.60	.09	.55
Tobacco/nicotine, γ702	−0.0017 (0.0111)	−0.16	.03	.88
Caffeine, γ703	0.0181 (0.0252)	0.72	.13	.48
Age, γ704	−0.0016 (0.0061)	−0.26	.04	.78
Wives' average CAR, γ800	0.0530 (0.0245)	2.16		.04
BMI, γ801	0.0037 (0.0045)	0.81	.13	.42
Birth control, γ802	0.1256 (0.0747)	1.67	.27	.10
Caffeine, γ803	−0.0001 (0.0281)	−0.01	.00	.99
Age, γ804	−0.0031 (0.0059)	−0.53	.09	.60

Note. None of the women in this sample reported tobacco or nicotine usage. Intercepts indicate average cortisol values at wakeup; average slopes of time since waking indicate change in cortisol per 1-hour change in time; average slopes of time since waking² indicate change in cortisol per 1-hour change in time²; CAR = Cortisol Awakening Response, indicating amount of change in cortisol during the 45 minutes after waking. Bolded rows indicate covariates that were included in subsequent analyses (p < .20).

Table 4.

Effects of Person-Level Work, Marital and Financial Predictors

Fixed effect	Coefficient (Standard Error)	T ratio	Effect Size r	p
Husbands’ intercept, γ100	1.0297 (0.0218)	47.21		.001
Hours per week that wife works, γ101	−0.0036 (0.0015)	−2.36	.37	.03
Hours per week that husband works, γ102	0.0005 (0.0014)	0.43	.07	.67
Length of marriage, γ103	0.0007 (0.0005)	1.54	.25	.13
Annual household income, γ104	−0.0232 (0.0221)	−1.05	.17	.30
Wives’ intercept, γ200	0.9669 (0.0283)	34.11		.001
Hours per week that wife works, γ201	−0.0045 (0.0020)	−2.22	.35	.03
Hours per week that husband works, γ202	0.0028 (0.0018)	1.55	.19	.13
Length of marriage, γ203	0.0014 (0.0006)	2.18	.34	.04
Annual household income, γ204	−0.0503 (0.0297)	−1.69	.27	.10
Husbands’ average slope of time since waking, γ300	−0.0785 (0.0049)	−15.90		.001
Hours per week that wife works, γ301	−0.0004 (0.0003)	−1.22	.20	.22
Hours per week that husband works, γ302	0.0003 (0.0003)	1.06	.17	.29
Length of marriage, γ303	−0.0001 (0.0001)	−1.81	.29	.07
Annual household income, γ304	−0.0000 (0.0052)	−0.01	.00	.99
Wives’ average slope of time since waking, γ400	−0.0699 (0.0055)	−12.64		.001
Hours per week that wife works, γ401	0.0013 (0.0003)	3.47	.50	.001
Hours per week that husband works, γ402	0.0004 (0.0004)	1.11	.18	.27
Length of marriage, γ403	−0.0001 (0.0001)	−1.48	.24	.14
Annual household income, γ404	0.0048 (0.0048)	1.00	.16	.32
Husbands’ average slope of time since waking2, γ500	0.0018 (0.0003)	5.90		.001
Hours per week that wife works, γ501	0.0000 (0.0000)	1.95	.31	.05
Hours per week that husband works, γ502	−0.0000 (0.0000)	−1.64	.26	.10
Length of marriage, γ503	0.0000 (0.0000)	1.60	.26	.11
Annual household income, γ504	0.0001 (0.0004)	0.53	.09	.59
Wives’ average slope of time since waking2, γ600	0.0016 (0.0004)	3.90		.001
Hours per week that wife works, γ601	−0.0001 (0.0000)	−2.11	.33	.04
Hours per week that husband works, γ602	−0.0000 (0.0000)	−1.46	.24	.15
Length of marriage, γ603	0.0000 (0.0000)	0.85	.14	.40
Annual household income, γ604	−0.0000 (0.0003)	0.13	.02	.90
Husbands' CAR, γ700
Hours per week that wife works, γ701	0.0025 (0.0018)	1.43	.23	.15
Hours per week that husband works, γ702	−0.0005 (0.0015)	−0.35	.06	.73
Length of marriage, γ703	−0.0008 (0.0005)	−1.54	.25	.12
Annual household income, γ704	0.0048 (0.0249)	0.19	.03	.85
Wives' CAR, γ800
Hours per week that wife works, γ801	0.0027 (0.0017)	1.58	.25	.11
Hours per week that husband works, γ802	0.0014 (0.0016)	0.91	.15	.36
Length of marriage, γ803	0.0001 (0.0006)	0.35	.06	.73
Annual household income, γ804	0.0107 (0.0246)	0.43	.07	.66

Note. Intercepts indicate average cortisol values at wakeup; average slopes of time since waking indicate change in cortisol per 1-hour change in time; average slopes of time since waking² indicate change in cortisol per 1-hour change in time²; CAR = Cortisol Awakening Response, indicating amount of change in cortisol during the 45 minutes after waking. Bolded rows indicate covariates that were included in subsequent analyses (p < .20).

Analysis to Determine Inclusion of Momentary NA as a Covariate

We next entered momentary NA as a Level 1 predictor (combination of Equations 1–4). As shown in Table 5, there was a significant partner effect of wives’ prior sampling-moment NA predicting husbands’ cortisol. Thus, when wives reported higher NA, their husbands showed higher cortisol levels approximately 1 hour to 1 ½ hours later on average. There were no other actor or partner affects of NA on cortisol that met our p < .20 cutoff for inclusion in subsequent analyses.

Table 5.

Effects of Momentary Negative Affect on Cortisol

Fixed effect	Coefficient (Standard Error)	T ratio	Effect Size r	p
Momentary effects on husbands' cortisol
Actor effects (effects of own NA)
Same-moment NA, π900	0.026 (0.231)	1.14	.19	.26
Prior-moment NA, π1000	0.020 (0.021)	0.95	.16	.34
Partner effects (effects of wives' NA)
Same-moment NA, π1100	−0.003 (0.021)	−0.13	.02	.90
Prior-moment NA, π1200	0.046 (0.019)	3.39	.49	.02
Momentary effects on wives' cortisol
Actor effects (effects of own NA)
Same-moment NA, π1300	0.018 (0.022)	0.80	.13	.42
Prior-moment NA, π1400	0.004 (0.020)	0.18	.03	.85
Partner effects (effects of husbands' NA)
Same-moment NA, π1500	0.006 (0.023)	0.26	.04	.78
Prior-moment NA, π1600	−0.003 (0.021)	−0.15	.03	.88

Note. Bolded rows indicate covariates that were included in subsequent analyses (p < .20). Additionally, all variables in bolded rows from Tables 2 through 4 are included as covariates in this analysis.

Hypothesis 1: Associations between Momentary Work Worries and Cortisol

We then tested the effects of momentary work worries on cortisol (combination of Equations 1–4, including the addition of wives’ prior-moment NA)². As shown in Table 6, there was an actor effect of husbands’ prior sampling-moment work worries predicting their own cortisol; husbands’ cortisol levels were unaffected by their wives’ work worries. Greater levels of wives’ cortisol were positively associated with their own same sampling-moment work worries (an actor effect) and independently associated with their husbands’ prior sampling-moment work worries (a partner effect). Thus, while husbands were physiologically reactive only to their own work worries, wives were physiologically reactive to their own work worries as well as to their husbands’ work worries.

Table 6.

Effects of Momentary Work Worries on Cortisol

Fixed effect	Coefficient (Standard Error)	T ratio	Effect Size r	p
Momentary effects on husbands' cortisol
Actor effects (effects of own work worries)
Same-moment work worries, π1000	−0.004 (0.019)	−0.19	.03	.85
Prior-moment work worries, π1100	0.048 (0.020)	2.44	.38	.02
Partner effects (effects of wives' work worries)
Same-moment work worries, π1200	0.016 (0.020)	0.80	.13	.42
Prior-moment work worries, π1300	−0.000 (0.019)	−0.00	.00	.99
Momentary effects on wives' cortisol
Actor effects (effects of own work worries)
Same-moment work worries, π1400	0.059 (0.022)	2.70	.41	.008
Prior-moment work worries, π1500	0.013 (0.021)	0.64	.11	.52
Partner effects (effects of husbands' work worries)
Same-moment work worries, π1600	−0.022 (0.020)	−1.07	.17	.29
Prior-moment work worries, π1700	0.053 (0.019)	2.80	.42	.006

Note. All variables in bolded rows from Tables 2 through 5 are included as covariates in this analysis.

The effects of work worries were moderated by the day of the week. As illustrated in Figure 3, for wives—but not husbands—the link between their own work worries and their own cortisol levels was stronger on Mondays compared to weekends (γ1410 = 0.097, SE = 0.032, t = 3.05, effect size r = .45, p = .003).

Figure 3.

Wives’ same-moment work worries predicting wives’ log-transformed cortisol, as moderated by Mondays vs. weekends. All variables in bolded rows from Tables 2 through 5 are included as covariates in this analysis.

Hypothesis 2: Moderation of Effects of Work Worries on Cortisol by Marital Disclosure

We next tested whether marital disclosure moderated the effects of momentary work worries on cortisol. This analysis was identical to the earlier model used to test the effects of same- and prior-sampling moment work worries on cortisol, but with marital disclosure and marital satisfaction and a marital disclosure X satisfaction interaction term added at Level 3; marital disclosure and marital satisfaction were both standardized prior to computing their interaction term. We found a significant two-way (cross-level) interaction between wives’ marital disclosure and their own same-moment work worries (γ₁₄₀₁ = −0.35, SE = 0.16, t = −2.29, effect size r = .36, p = .022), indicating that wives who reported low levels of marital disclosure showed a stronger positive association between work worries and cortisol than those who were highly disclosing. The interaction between wives’ marital satisfaction and work worries was not significant (γ₁₄₀₂ = −0.17, SE = 0.11, t = −1.60, effect size r = .26, p = .11), but there was a significant marital disclosure X satisfaction X work worries three-way interaction (γ₁₄₀₃ = .49, SE = 0.18, t = 2.65, effect size r = .40, p = .009). As shown in Figure 4, for those wives who indicated that they were either highly disclosing or highly satisfied with their spouses, there was essentially no association between momentary work worries and cortisol (and no significant differences between the slopes). However, for wives who indicated that they were both low in marital disclosure and low in marital satisfaction, there was a strong positive association between work worries and cortisol (simple slope = .56, SE = .14, t = 4.03, effect size r = .56, p < .0003). In other words, wives who were unsatisfied in their relationship and were not open about their thoughts and feelings with their husbands were very physiologically reactive to their own work worries. For husbands, neither marital disclosure nor marital satisfaction moderated the association between momentary work worries and cortisol.

Figure 4.

Wives’ same-moment work worries predicting wives’ cortisol, as moderated by wives’ wives’ marital satisfaction and marital disclosure, with high and low values plotted at +1 standard deviation and −1 standard deviation from the mean, respectively (54). All variables in bolded rows from Tables 2 through 5 are included as covariates in this analysis.

Discussion

We found that momentary changes in work worries had a physiological impact on both the person experiencing the worry as well as on the person’s spouse. Greater work worries were related to one’s own cortisol levels for both husbands and wives—actor effects. Husbands’ work worries also independently predicted wives’ cortisol levels—a partner effect. Among wives, lower marital disclosure augmented the association between momentary work worries and cortisol levels, with wives who were both low in disclosure and low in marital satisfaction showing a strong positive relationship between work worries and cortisol.

Our work worry findings are novel in several respects. First, although prior EMA studies have investigated the links between mood and cortisol in everyday life (8, 41, 42), none to our knowledge has examined the effects of specific types of worries. It is notable that neither negative affect nor daily levels of perceived stress were associated with cortisol after taking into account the effects of stressful feelings about work. A number of studies have shown that stressful experiences earlier in the day at work can spillover into everyday family life and negatively influence mood at home (3, 43, 44). Our findings suggest that worries about work when employed individuals are home with their families can affect physiology as well.

Second, this is the first study to our knowledge to demonstrate that one person’s momentary feelings of stress are related to another person’s cortisol levels in daily life. Husbands’ work worries affected wives cortisol levels, but wives’ feelings of work stress did not affect husbands’ cortisol levels. This is in line with previous laboratory research suggesting that wives are more physiologically reactive to marital interactions than husbands (45, 46) as well as naturalistic work showing that couples’ cortisol levels covary (47). While previous research has shown effects of interpersonal stress on physiology, these findings suggest that for women at least, partner effects on physiology also extend to non-relationship stressors.

By incorporating both same and prior sampling-moment psychological measures, we were able to explore the timing of the links between momentary work worries and cortisol in a naturalistic setting. The differences in the time course of stress-cortisol links for husbands and wives—with husbands’ cortisol levels influenced by prior-moment work worries and wives’ cortisol levels influenced by same-moment work worries—were unexpected. In laboratory studies, acute stressors typically lead an increase in salivary cortisol levels in approximately 20 minutes (48). At home, when mood can change over time depending on situational cues and other factors, the timing of psychological states on cortisol is likely to be more variable. Unlike the lab, in everyday life, measuring the exact onset and offset of stressors is difficult. Further, there may be systematic differences in men’s and women’s cognitive and emotional processing of stress that lead to a greater time lag among men than women.

In addition to momentary effects of work worries on cortisol, we found person-level effects of marital disclosure moderating the links between momentary work worries and cortisol among wives: wives who reported low levels of marital disclosure showed a much stronger association between work worries and cortisol than those who were highly disclosing. This effect was especially pronounced among wives reporting both low marital disclosure and low marital satisfaction. Wives who were either disclosing to their husbands or satisfied in their marriages did not appear to be physiologically reactive to their own work worries.

The marital disclosure findings reported here suggest that for wives at least, the stress-buffering effects of a happy marriage on HPA activity (4) may be partially explained by the extent to which they are able to be open and expressive with their spouses. The benefits of written emotional disclosure for health have been documented (17). This finding suggests that disclosure in the context of close relationships—the context that affords the greatest opportunities for self-disclosure—is beneficial as well.

There are limitations of this study that should be noted. The first is statistical power. Using Cohen’s criteria of r = .10 for small effects, r = .30 for medium effects and r = .50 for large effects (49), post hoc power analyses (23) showed that we had very little power for detecting small effects (power = .13), reasonable power for detecting medium effects (power = .74) and excellent power for detecting large effects (power = .90). Attempts at replication with larger probability samples would determine the robustness of these findings and evaluate their generalizability. For example, it would be worthwhile to investigate whether the effects reported here vary as a function of stage of life and an absence of children in the home. It may be that the effects of work-related stressors are especially strong during the early years of parenthood and one’s career compared to later years.

Second, the uneven spacing of sampling moments throughout each day was not ideal for the type of lag momentary analyses conducted here. Future studies should consider additional assessments (e.g., 60 and 20 minutes prior to each saliva-sampling occasion) in order to more precisely estimate the time course of the effects of everyday feelings of stress on cortisol. Of course, the tradeoff of additional participant burden in this type of design is also a consideration.

Despite these limitations, this study represents an important step in furthering our understanding of how everyday feelings of stress influence one’s own physiology and the physiology of close others. We were able to test simultaneously the within-couple effect of momentary work worries and the between-couple stress-buffering effect of marital disclosure on cortisol. HPA axis dysregulation has been linked to negative health consequences—including mortality (50)—through wear and tear brought on by chronic stress (51). The findings reported here indicate that taking into consideration the perspectives of close others will be vital to our understanding of daily stress processes. Although a growing number of researchers are going beyond the individual to examine the impact of the quality of relationships—especially marriage—on health (52, 53), very little is known about the precise aspects of relationships that confer health benefits. Examining specific relationship processes such as marital disclosure and their buffering effects on stress-health associations ultimately will lead to a deeper and more complex understanding of the health benefits of social relationships.

ABBREVIATIONS

HPA

hypothalamic-pituitary-adrenal

EMA

Ecological Momentary Assessment

MAT

Marital Adjustment Test

NA

negative affect

PSS

Perceived Stress Scale

APIM

Actor-Partner Interdependence Model

CAR

cortisol awakening response