Covariation in Couples’ Nightly Sleep and Gender Differences

Abstract

Objectives

For most partnered adults, sleep is not an individual-level behavior―it is a shared health behavior with a partner. This study examined whether perceived nightly sleep duration and sleep quality covaried within couples and whether the unique influence of partner sleep on individual sleep differed by gender.

Design

Eight consecutive days of diary data.

Participants

U.S. hotel employees and their spouses/partners (N=76 from 38 couples, 600 daily observations)

Measurements

Each day, couples separately reported their previous night’s sleep duration (in hours) and sleep quality (1=very unsatisfactory to 5=very satisfactory). Analyses adjusted for sociodemographic, family, work, and day-level characteristics.

Results

Dyadic multilevel modeling revealed positive covariation in nightly sleep duration within couples. After controlling for the effects of contextual covariates, partner influence on individual sleep duration was more apparent in men’s sleep. When a female’s sleep duration was longer or shorter than usual, their male partner’s sleep duration was also longer or shorter than usual, respectively. However, a female’s sleep was not significantly predicted by her male partner’s sleep duration, after taking into account the effects of her sleep on the male partner’s sleep and contextual covariates. Sleep quality covaried on average across days between partners, and this association did not differ by gender.

Conclusions

Our results demonstrate positive covariation in sleep duration and sleep quality within couples. Couples’ sleep duration covaried night-to-night and their sleep quality covaried on average across days. A male’s sleep duration is predicted by the female partner’s sleep duration, but not vice versa. Future research should examine health consequences of couple sleep covariation.

Introduction

Sleep is a biological need also influenced by social contexts. One such social context is the couple relationship, where an individual’s sleep can be influenced by the paired partner’s sleep. Yet most research has assessed sleep as an individual-level behavior, limiting the ability to understand the dyadic nature of sleep.¹ Emphasizing the importance of dyadic relationships, previous research documents health concordance within couples. For example, a systematic review of 103 studies² shows similarity of mental and physical health within couples. There is also evidence that an individual’s momentary affect and physiology positively covary with the partner’s momentary affect and physiology.³ These studies suggest that couples may also covary in their sleep characteristics. The purpose of this study is to examine whether and how sleep duration and sleep quality covary within couples across multiple diary days.

The shared resource hypothesis⁴ suggests that married or partnered couples share a pool of resources and environment. Couples also tend to share temporal resources on a daily basis, such as spending leisure time together.⁵ Shared environment and time may translate into shared health risks or benefits within couples. For example, if one spouse smokes, the other spouse is more likely to be a smoker or exposed to secondhand smoke.⁶ This suggests that if one spouse gets less sleep, the other spouse may also be more likely to have less sleep. Consistent with this view, a small but growing line of research has shown a positive covariation in sleep-wake timing (or “sleep concordance”) within couples measured via actigraphy.^7,8 Actigraphy has a number of benefits, such as the ability to objectively measure sleep,⁹ however, it does not provide participants’ subjective assessments of their sleep. Using dyadic daily diary, the current study tests whether couples covary in their perceived sleep duration and sleep quality. Building on the prior work on sleep concordance based on actigraphic sleep timing,^7,8 we expect to find a positive covariation in nightly perceived sleep duration and sleep quality within couples. We further assess whether an individual’s sleep covaries with their partner’s sleep, averaged across multiple diary days, or on a day-to-day basis. Multilevel sleep covariation has not been examined in prior research, but is important to understand the specific nature of covariation that may differ by sleep duration and quality.

Moving beyond sleep covariation within couples, this study also tests gender differences in sleep covariation. Although sleep covariation may be a by-product of couples’ shared environments,⁴ there may be gender differences in how individuals’ sleep uniquely predicts or is predicted by their partner’s sleep beyond the effects of contextual factors. Women, in general, experience more sleep disruption than men due to work-family responsibilities.¹⁰ Women’s frequent role as a primary caregiver at home can generate more causes for their sleep disruption.¹¹ For example, the presence of a young child may be a more important predictor of a female’s sleep than her male partner’s sleep. For working women, their sleep may also be more sensitive to differences in demands by daily contexts, such as workday vs. non-workday or weekdays vs. weekends, rather than to their partner’s sleep. Moreover, Umberson’s social control perspective¹² suggests that wives are more likely to attempt to influence the health behaviors of their husbands. As women’s socialization and experiences are more oriented toward nurturance and caring for others than are men’s, women are more likely to exert control over others’ health behaviors, especially those of their spouses/partners. Indeed, several studies have found stronger and more successful social control strategies by women than by men, and that male partners’ health behaviors (e.g., diet, exercise, smoking cessation) are more affected by their female partner’s social control than the reverse.^13,14 Taken together, the social control perspective¹² suggests that after controlling for the effects of contextual factors on individual sleep, a male’s sleep may be predicted by the female partner’s sleep, but not vice versa. In this study, we consider sociodemographic, family, work, and day characteristics as contextual factors found important for individual sleep.^11,15

Present Study

Using daily dyadic data collected from male and female partners on the same nights, this study evaluates the shared nature of nightly sleep in a sample of couples in midlife. Our daily diary approach enhances ecological validity by examining couples’ sleep in their everyday life.¹⁶ Multiple days of data from both male and female partners also allow us to test how sleep covaries within couples over a week. Figure 1 shows our research model. We first tested whether there is positive covariation in couples’ diary reports of sleep in terms of duration and quality in a multilevel framework. In testing this, we assessed whether an individual’s sleep covaried with their partner’s sleep, on average across days (at the between-person level), or on a daily basis (at the within-person level). Then we examined gender differences in the influence of partner sleep, beyond the influence of contextual factors on their sleep. To examine this, we adopted dyadic multilevel modeling.^17–19 This method not only accounts for statistical interdependence within couples but also gives higher confidence in examining the effect of female partner’s sleep on the paired male partner’s sleep, after taking into account the effect of male partner’s sleep on the female partner’s sleep, and vice versa. We hypothesized that, beyond the effects of contextual factors on individual sleep, men’s sleep would be predicted by their female partner’s sleep, but not vice versa. In addition, we further explored whether relationship quality moderated partner influence on individual sleep, given the importance of relationship quality in couple relationships and sleep.^7,20–23

Figure 1.

Research model testing multilevel couple sleep covariation and gender differences

Note. White boxes indicate covariates; grey boxes indicate predictors; black boxes indicate outcomes.

Method

Data for the current analyses are from the Hotel Work and Well-Being Study,²⁴ a project investigating connections between work stress, health, and family relations of employees from the hotel industry. Comprehensive details of the design and sample can be found in previous research,²⁵ with details relevant to the current analysis provided below.

Participants and Procedure

Following a baseline telephone survey, hotel managers who were married or in a cohabiting relationship were invited to participate in a daily diary study together with their spouse/partner. Among 98 hotel managers who participated in the daily diary study, 42 spouses/partners also participated in the daily diary study. These 42 couples were telephoned on eight consecutive nights and separately asked about their daily experiences, including sleep duration and sleep quality. Of these, 4 couples were same sex couples and thus excluded from the current analyses that focused on gender differences in dyadic sleep among heterosexual couples. Thus, our final analytic sample consisted of 38 heterosexual couples (76 individuals). Participants received a $50 gift card for their participation. This study was approved by appropriate Institutional Review Boards and conducted in accordance with the Declaration of Helsinki.

Male partners’ mean age was 37.2 (SD = 8.9, Range = 24–64); 73% were White, 11% were Hispanic or Latino, 8% were African American, and 8% were Southeast Asian or Asian American; 63% had completed four years of college or more and 29% had some college (1–3 years) or technical school; most (92%) were employed and 57% worked standard shifts (vs. nonstandard shifts). Female partners’ mean age was 35.0 (SD = 8.4, Range = 24–61); 68% were White, 18% were Hispanic or Latino, 11% were African American, and 3% were Southeast Asian or Asian American; 61% had completed four years of college or more and 24% had some college (1–3 years) or technical school; the majority (87%) were employed and 58% worked standard shifts (vs. nonstandard shifts). Ninety-seven percent of the couples were married and the rest were cohabiting with a permanent romantic partner. For those who were married, mean years in marriage was 8.92 (SD = 9.12). Fifty-five percent had a child. For those with any children, their mean number of children was 2.67 (SD = 1.39) and 52% had children younger than 6 years.

Measures

Nightly sleep duration

Each evening, respondents were asked “Since this time yesterday, how much time did you spend sleeping, not including time you may have spent napping?” Respondents provided hours and minutes, and nightly sleep duration was calculated. The intra-class correlation (ICC)²⁶ indicated that, of the total variance in sleep duration pooled across men and women, 3% was due to between-couple differences, 23% was due to between-person differences, and 74% was attributable to day-to-day fluctuations within persons. When variance components were examined separately for men and women, about 25% of the total variance was due to between-person differences and the remainder was attributable to within-person fluctuation (Figure 2).

Figure 2.

Variance components in sleep duration and sleep quality.

Nightly sleep quality

Each evening, respondents were asked “How would you describe your sleep (since this time yesterday)?” Responses ranged from 1 (very unsatisfactory) to 5 (very satisfactory), with higher scores indicating higher sleep quality. The ICC of sleep quality showed that 15% of the total variance was due to between-couple differences, 16% was due to between-person differences and 69% was attributable to within-person fluctuations. For men, 44% of the total variance was due to between-person differences and 56% was attributable to day-to-day variations. For women, 20% of the total variance was due to between-person differences and 79% was attributable to day-to-day variations (Figure 2).

Covariates

We included age (in years, centered at the sample mean), race (white vs. non-white), education (less than college degree vs. college degree or more), and presence of young children (age < 6) in household (vs. not) as covariates. We also took into account potential differences in sleep due to non-paid work (=1) or working nonstandard hours (=2), both compared to working standard hours (=3; reference category). The types of nonstandard shifts varied with no dominant type (e.g., evening shift: 9%, night shift: 3%, rotating shift: 3%, split shift: 4%, working additional hours in the evening or on weekends; 19%, and other; 4%), thus we combined all different types of nonstandard shifts into one category. Moreover, we controlled for day effects, including weekend (vs. weekdays) and non-work days (vs. workdays).

Relationship quality

Additionally, to explore the potential moderating role of overall relationship quality in couple sleep covariation, we assessed the global measures of partner support and partner strain, collected during the baseline telephone survey about one month prior to the daily diary study. The two scales were used and validated in the Midlife in the United States (MIDUS) study.^27,28 The partner support scale contained seven items, such as “How much does your spouse or partner really care about you?” The partner strain scale had six items, such as “How much do you feel your spouse or partner makes too many demands on you?” Participants answered on a 4-point scale from 1 (not at all) to 4 (a lot). Item scores in each scale were summed for each individual. Cronbach’s alpha for partner support was .66 and the alpha for partner strain was .70. The correlation between partner support and partner strain was −0.40 (p < .001), suggesting that they are two distinct variables.

Data Analysis

We used dyadic multilevel modeling in SAS 9.4 to account for statistical interdependence within couples.^17–19 Models used stacked data where male and female partners’ data were on separate lines and nested within couple-level IDs. Most couples completed all 8 days’ diaries; 2 couples completed 7 days and 1 couple completed 6 days, resulting in 600 daily observations nested within 38 couples. Of the 600 total diary days, 92% had valid responses on sleep questions. Within-person (level 1) and between-person (level 2) levels of partner’s sleep were entered as predictors of own sleep. Within-person variables were centered at the person mean, such that positive values indicate values higher than the person’s (partner’s) own cross-day average. Between-person variables were centered at the sample mean, such that positive values indicate higher scores than the sample average. The between-person effect indicates, for example, whether an individual whose partner report longer sleep than the sample average also reports longer sleep across days. The within-person effect of partner sleep on the other partner’s sleep indicates, for example, on nights when partner “A” reports sleeping longer than his usual sleep duration, whether partner “B” also reports sleeping longer than her usual. We separately modeled sleep duration and sleep quality in each multilevel model.

To test our hypothesis examining gender differences, separate intercept and slope terms were created for men and women, with partners denoted by dummy variables. This technique has been commonly adopted in dyadic analysis,^17–19 which allowed us to separately and simultaneously predict men’s and women’s sleep in a single model. For example, the level 1 model for sleep duration (without covariates) was specified as:

$${\text{Own Sleep Duration}}_{\text{di}}={\mathrm{\beta }}_{\text{0i}}{\text{MALE + β}}_{\text{0i}}{\text{FEMALE + β}}_{\text{1i}}{\text{MALE×Partner Sleep Duration}}_{\text{di}}{\text{+ β}}_{\text{1i}}{\text{FEMALE×Partner Sleep Duration}}_{\text{di}}+{\mathrm{\epsilon }}_{\text{ti}}$$

MALE and FEMALE are dummy coded indicators for men and women. These two dummy codes indicate whose outcome to estimate. For example, β_1iMALE represents the predictive effect of a female partner’s sleep duration on the paired male partner’s sleep duration. The model allowed random intercepts for men and women, such that each man and woman had their own intercept, which also took into account non-independence in dyad members’ scores. However, the model did not allow random slopes due to the lack of lower-level units within a couple to allow the slopes to vary from couple to couple.¹⁹

Results

Table 1 shows between-person level unadjusted descriptives and within-couple correlations for sleep variables. For men, mean sleep duration was 6.77 hours per day, and mean sleep quality was 3.95 which corresponds to “satisfactory.” For women, mean sleep duration was 6.95 hours per day, and mean sleep quality was 3.84. There were no significant differences in sleep duration and quality between partners. Male partners’ nightly sleep duration and sleep quality were significantly and positively correlated with the paired female partner’s nightly sleep duration and sleep quality.

Table 1.

Descriptives and within-couple correlations for key variables

Group and variable	M	SD	Range	Difference between partners, t-test	Correlation with partner, r
Men (n = 38)
Sleep duration	6.77	0.88	4.63 to 8.69	0.85	0.14*
Sleep quality	3.95	0.64	2.43 to 5.00	−0.77	0.22***
Women (n = 38)
Sleep duration	6.95	0.91	4.80 to 8.25
Sleep quality	3.84	0.54	2.80 to 5.00

Note. Means, Standard Deviations, and Ranges were based on person means across days. The within-couple correlations for sleep duration and sleep quality were estimated at the daily level (n_obs = 261, 260, respectively).

^*p < .05,

^**p < .01,

^*p < .001.

Multilevel Covariation in Sleep Duration and Sleep Quality within Couples

Then we examined the specific nature of couple sleep covariation in a multilevel framework. We tested the between-person and within-person effects of partner sleep on one’s own sleep, using pooled effects across men and women (Table 2). Beginning with sleep duration, individuals who had less education (less than 4 years of college) had shorter sleep than those with more education (college degree or more). Days with non-paid work were associated with longer sleep than days with paid work. After controlling for these effects of covariates, there was a significant within-person effect of partner sleep duration on own sleep duration. That is, on days when partner “A” slept longer than usual, the paired partner “B” also slept longer than usual. There was no significant between-person effect of partner sleep duration. That is, an individual’s sleep duration did not significantly differ from others in the sample depending on partners’ average sleep duration.

Table 2.

Multilevel effects of partner sleep on individual’s own sleep

	Sleep duration (in hours)		Sleep quality (1=very unsatisfactory to 5=very satisfactory)
	B	(SE)	B	(SE)
Fixed effects
Intercept	6.63 ***	(0.26)	3.68 ***	(0.16)
Sociodemographic, Family, and Work characteristics
Gender, Men (vs. Women)	−0.25	(0.22)	0.07	(0.13)
Age	0.00	(0.01)	0.00	(0.01)
Race, White (vs. non-White)	0.28	(0.24)	0.23	(0.14)
Education, Some college or below (vs. College grad or above)	−0.66 **	(0.25)	−0.18	(0.15)
Presence of young children in household (Age < 6)	−0.22	(0.25)	−0.17	(0.15)
Not working (vs. work standard shifts)	−0.35	(0.41)	0.15	(0.24)
Working nonstandard shifts (vs. work standard shifts)	0.04	(0.24)	0.07	(0.14)
Weekend (vs. Weekdays)	0.19	(0.14)	0.25**	(0.09)
Non-workday (vs. Workday)	0.91 ***	(0.16)	0.18 †	(0.10)
Partner’s Sleep
Between-Person level Partner Sleep	0.03	(0.12)	0.32 **	(0.11)
Within-person level Partner Sleep	0.13 **	(0.05)	0.05	(0.05)
Random effects
Between-men variance	0.60 **	(0.20)	0.33 **	(0.11)
Between-women variance	0.65 ***	(0.21)	0.14 **	(0.06)
Residual variance	1.36 ***	(0.09)	0.54 ***	(0.04)

Note. 600 observations were nested within 38 couples. 514 and 512 observations were used in the models for sleep duration and sleep quality, respectively, due to missing responses in variables. Main predictors were capitalized.

^†p < .10,

^*p < .05,

^**p < .01,

^***p < .001

Turning to sleep quality, there was a significant difference between weekends vs. weekdays. Individuals’ sleep quality was higher on weekends than on weekdays. In addition, partner sleep quality significantly predicted own sleep quality at the between-person level. Individuals who lived with a partner whose average sleep quality was higher also had higher sleep quality on average across days than those with a partner whose average sleep quality was lower. There was no significant within-person effect of partner sleep quality on one’s own nightly sleep quality. Thus, we found that couples’ sleep duration covaried night-to-night, and their sleep quality covaried on average across days.

Unique Effects of Partner Sleep on Men’s and Women’s Sleep, Beyond Contextual Factors

Table 3 shows results of multilevel models simultaneously predicting men’s and women’s sleep duration and sleep quality. With regard to men’s sleep duration, after taking into account the significant effects of education and non-work days, within-person partner sleep duration was a significant predictor: On days when a female partner slept longer or shorter than her usual, the male partner also slept longer or shorter than his usual. For women, this within-person partner effect was not significant after controlling for the contextual covariates. These effects were estimated simultaneously in the same model, and hence the results indicate the unique effect of a female partner’s sleep on the male partner’s sleep duration, independent of the effect of a male partner’s sleep on the female partner’s sleep duration on the same night.

Table 3.

Partner sleep separately and simultaneously predicting men’s and women’s sleep

	Sleep duration (in hours)		Sleep quality (1=very unsatisfactory to 5=very satisfactory)
	B	(SE)	B	(SE)
Fixed effects
Predicting Men’s Sleep
Intercept	6.71 ***	(0.36)	3.90 ***	(0.23)
Men’s Sociodemographic, Family, and Work Characteristics
Age	0.02	(0.02)	0.01	(0.01)
Race, White (vs. non-White)	0.18	(0.36)	−0.27	(0.23)
Education, Some college or below (vs. College grad or above)	−0.93 *	(0.35)	0.01	(0.25)
Presence of young child (vs. not)	−0.29	(0.36)	−0.14	(0.24)
Not working (vs. work standard shifts)	−0.46	(0.63)	−0.44	(0.40)
Working nonstandard shifts (vs. work standard shifts)	0.03	(0.34)	0.47 *	(0.22)
Weekend (vs. Weekdays)	0.11	(0.20)	0.33 **	(0.12)
Non-workday (vs. Workday)	0.87 ***	(0.22)	0.10	(0.14)
Female Partner’s Sleep
Between-Person level Partner Sleep	0.08	(0.18)	0.46 *	(0.22)
Within-person level Partner Sleep	0.16 *	(0.06)	0.06	(0.06)
Predicting Women’s Sleep
Intercept	6.55 ***	(0.41)	3.62 ***	(0.20)
Women’s Sociodemographic, Family, and Work characteristics
Age	−0.02	(0.02)	0.01	(0.01)
Race, White (vs. non-White)	0.29	(0.37)	0.35†	(0.19)
Education, Some college or below (vs. College grad or above)	−0.26	(0.37)	−0.28	(0.18)
Presence of young child (vs. not)	−0.14	(0.38)	−0.18	(0.19)
Not working (vs. work standard shifts)	−0.54	(0.60)	0.37	(0.30)
Working nonstandard shifts (vs. work standard shifts)	−0.02	(0.38)	−0.05	(0.19)
Weekend (vs. Weekdays)	0.25	(0.21)	0.18	(0.13)
Non-workday (vs. Workday)	0.97 ***	(0.24)	0.25†	(0.15)
Male Partner’s Sleep
Between-Person level Partner Sleep	0.07	(0.18)	0.32 *	(0.13)
Within-person level Partner Sleep	0.10	(0.07)	0.05	(0.08)
Random effects
Between-men variance	0.63 **	(0.22)	0.25**	(0.09)
Between-women variance	0.69 **	(0.23)	0.14 **	(0.06)
Residual variance	1.37 ***	(0.09)	0.54 ***	(0.04)

Note. 600 observations were nested within 38 couples. 514 and 512 observations were used in the models for sleep duration and sleep quality, respectively, due to missing responses in variables. Main predictors were capitalized.

^†p < .10,

^*p < .05,

^**p < .01,

^***p < .001

In terms of sleep quality, for both men and women, between-person partner sleep quality significantly predicted their own sleep quality. For men’s sleep quality, their nonstandard shifts (vs. standard shifts) and weekends (vs. weekdays) were significant covariates. For example, the effect of nonstandard shifts on men’s sleep quality indicated that, when men worked nonstandard shifts, their sleep quality was 0.5 units higher (on a 5-point scale) compared to when they worked standard shifts. This effect remained after adjusting for the effects of all other predictors of both men’s and women’s sleep quality simultaneously included in the model. The effect of male partners’ nonstandard work shifts on their sleep quality was thus independent of the degree that the paired female partner’s work shifts predicted her sleep quality. There was no gender difference in the covariation in sleep quality. Thus, our hypotheses that suggest, after controlling for the effects of contextual factors, partner’s sleep would uniquely predict men’s sleep (but not women’s sleep) was supported in terms of sleep duration, but not in sleep quality.

Potential Moderating Role of Relationship Quality

In addition, we explored whether relationship quality moderated the effects of partner sleep on individual sleep. There was a significant interaction with partner support predicting women’s nightly sleep duration (Female: Within-person partner sleep duration × Own perceived partner support, B = −0.11, SE = 0.04, p < .01). To interpret this interaction effect, we dichotomized the level of partner support using the mean split (M=22.63, Range=18–24). Scores above the mean (61%) were coded as higher partner support, and scores at the mean or below were coded as lower support (39%). Figure 3 shows that, for a female who perceived lower partner support, the degree of covariation with the male partner’s nightly sleep duration was positive and significant. For a female who perceived higher partner support, the degree of covariation was not significant. These effects were not found in men. There was no significant interaction with partner strain either for women or men. There were no moderating effects of partner support or strain in terms of covariation in sleep quality.

Figure 3.

The role of women’s perceived partner support on within-couple covariation in nightly sleep duration.

Note. For female partners who perceived lower partner support, the degree of covariation with their male partner’s sleep duration was positive and significant (B = 0.39, SE = 0.12, p < .01). For female partners who perceived higher partner support, the degree of covariation was not significant (B = −0.02, SE = 0.08, p > .10). The model adjusted for the effects of sociodemographic, family, work, and daily contextual covariates.

Discussion

This study investigated sleep covariation within couples. Our daily diary design allowed for examining couple sleep covariation at multiple levels that may differ by sleep measures. We found that sleep duration covaried within couples on a daily basis (within-person level) and sleep quality covaried on average across days (between-person level). Moreover, we tested gender differences in sleep covariation beyond the influence of contextual factors. Building on prior research regarding gender differences in social roles and social control in couple relationships,^10–12 we hypothesized that a male’s sleep would be predicted by the female partner’s sleep (but not vice versa). Our hypothesis was supported for sleep duration, but not for sleep quality. Our findings shed light on the dyadic nature of sleep within couples and underscore the importance of considering partner’s sleep when examining individual sleep in partnered adults.

Night-to-night covariation in sleep duration and average covariation in sleep quality within couples

This study advances prior literature examining within-couple concordance in sleep timing,^7,8,23 by showing positive within-couple covariation in sleep duration and sleep quality evident in couples’ daily diary reports. The results also provide a new knowledge on couple sleep by revealing the different nature of covariation between sleep duration and sleep quality. Sleep duration covaried within couples night-to-night, however, did not covary on average across multiple diary days. This indicates that, within a couple’s relationship, how long one partner sleeps on a night has a proximal effect on how long the other partner sleeps on that night. However, one partner’s habitual sleep duration does not have an overall effect on the other partner’s habitual sleep duration over a week. In contrast, sleep quality covaried on average across days, but not on a day-to-day basis. Living with a partner who has higher average sleep quality was associated with higher average sleep quality. This finding resonates with the effect of shared environment⁴ on couple-level overall sleep quality. If couples share an environment that promotes good sleep behaviors (e.g., lack of screens or light in the bedroom), they may perceive a better sleep quality on average. Future research should examine factors related to a shared environment that promote or inhibit couple-level sleep quality.

A male partner’s nightly sleep duration is predicted by the female partner’s nightly sleep duration, but not vice versa

Although there was an overall effect of partner sleep duration predicting own sleep duration at the within-person level, the unique influence of partner sleep on individual sleep was more apparent for men’s sleep duration. Specifically, after taking into account the effects of contextual covariates, a male partner reported that he slept longer than his usual on days when the female partner slept longer than her usual, rather than vice versa. This finding is in line with the social control perspective which suggests a strong influence of women on their partner’s health.¹² In contrast, a female’s sleep duration was not significantly predicted by the male partner’s sleep duration after taking into account contextual factors. This suggests that a female’s sleep duration may be more affected by daily contexts or demands (e.g., non-workday as found in this study), rather than by her male partner’s sleep duration. There may be other important predictors of women’s sleep duration that the current study did not test, such as children’s sleep.^10,11 In terms of sleep quality, however, there was no gender difference in the influence of partner sleep. Again, sleep quality did not significantly covary on a daily basis. It may be that men and women differ in their daily evaluation of sleep quality, but they are similar in overall sleep quality. More studies are needed to understand the daily process of couple sleep covariation, such as whether and how daily interactions between partners and momentary moods before bed contribute to couple sleep covariation.

A female partner who perceives lower partner support tended to have sleep quantity similar to the male partner’s

Our further exploration revealed that a female’s sleep duration was predicted by the male partner’s sleep duration only when she perceived lower partner support, but not when she perceived higher partner support. This result is intriguing that female partners in lower quality relationships were more (not less) affected by their male partner’s nightly sleep duration. However, a similar pattern was reported in previous studies examining couple-level covariation in physiology: the degree of covariation in cortisol was stronger for couples with low relationship quality.^3,29–31 Perhaps, women who perceived less support in their relationships might have adjusted their sleep to their partner’s sleep, given that women in heterosexual couple relationships often report that they invest more effort than men in their relationships.³² Another possible explanation is that women who perceived less partner support might have had less sleep than those with more support; as a result, their sleep duration becomes more similar to that of men who tend to have shorter sleep than women in this sample (see Table 1). This may relate to a potential issue of “dual sleep deficiency” in couples. As depicted in Figure 3, when their male partner had less than 7 hours of sleep, female partners had less than 6 hours of sleep exhibiting a linear decrease with every 30 minutes decreases in the male partner’s sleep. However, female partners who perceived higher levels of partner support did not exhibit a significant effect of their male partner’s sleep duration on their own sleep duration. More research is needed to replicate and understand this intriguing finding.

Limitations and Future Directions

Despite this study’s contributions, some limitations suggest directions for future research. First, although our sample of couples lived in the same household, we do not know whether they shared a bed night to night. However, if non-bedsharing couples were included, this would lead to an underestimation of the degree of couple sleep covariation, thus making our estimate more conservative. Future diary studies could ask about couples’ bed sharing with consideration of several factors (e.g., the number of hours that meet criteria for sharing a bed, and which partner’s report is most important)³³ to better understand the degree of sleep covariation within bedsharing couples. Second, the study sample was purposively selected from the hotel industry (i.e., one partner worked at a hotel), and thus our findings may not generalize to couples in other contexts. Considering that most hotel employees were hourly workers²⁵ and approximately half worked nonstandard work schedules (although there was no significant difference in sleep duration by work shifts in this study), we may have underestimated the degree of within-couple covariation in nightly sleep duration in the general population. Future research should examine couple-level sleep covariation in diverse samples. Third, this study used self-reports to measure nightly sleep duration and sleep quality. Although our focus was to examine within-couple covariation in perceived sleep duration and sleep quality measured by daily diary reports, future research could benefit from incorporating actigraphically-assessed sleep duration and sleep quality (e.g., wake after sleep onset, sleep fragmentation). In addition, our participants completed the daily diary each evening, which might have caused a recall bias in reporting previous night’s sleep duration and sleep quality relative to the more typical morning sleep diary. Further, information on daytime napping was not collected. Future studies could consider assessing couples’ daily sleep patterns more extensively, such as asking previous night’s perceived sleep in the morning and measuring daytime napping as well as nighttime sleep via actigraphy methods. Lastly, our assessment of nightly sleep did not include more specific information regarding sleep problems, such as delayed sleep onset, difficulty maintaining sleep, or obstructive sleep apnea (OSA). Previous research reported that OSA has a significant influence on partner’s sleep.¹ In future research, it would be valuable to consider the potential role of these sleep problems and other medical disorders in individual sleep quantity and quality and also in the degree of covariation with their partner’s sleep.

Conclusions

This study reports that individuals’ sleep is likely to be predicted by their partner’s sleep in couple relationships. Couples’ sleep duration positively covaried night-to-night and their sleep quality positively covaried on average across days. Strengths of this study include a dyadic daily diary design, which sampled multiple assessments separately from each partner with ecological validity, and an analytic approach that tested multilevel effects of partner’s sleep on men’s and women’s sleep simultaneously. At the most general level, our findings suggest that partner influence should be considered when examining sleep characteristics among partnered adults. By understanding the shared nature of sleep and who’s sleep is more affected by whose sleep and under what circumstances, we will be able to better target interventions to improve individual and family sleep.