The Role of Couple Sleep Concordance in Sleep Quality: Attachment as a Moderator of Associations

Summary

Despite most American adults sharing a bed with a romantic partner, sleep research has examined sleep primarily as an individual behavior. A growing body of research indicates that couple bed sharing may have an impact on sleep quality, but the current study is the first to examine whether such associations may differ based on attachment security. A sample of 179 cohabiting heterosexual couples completed daily sleep diaries and surveys of their attachment security, avoidance, and anxiety. Data were analyzed using multilevel modeling. Greater attachment security and lower attachment avoidance were associated with greater subjective sleep quality. Greater sleep concordance (time in bed with partners) was associated with better subjective sleep quality for women with lower attachment security and higher attachment avoidance. Findings suggest that couple bed sharing may benefit the subjective sleep quality of women who have lower attachment security.

Introduction

Subjective sleep quality is associated with physical and mental health, emotion regulation, cognitive performance, well-being, and life satisfaction (Tavernier & Willoughby, 2015; Moore et al., 2002; Pilcher & Ott, 1998). Thus, understanding factors that promote or undermine sleep quality is important. Research has only recently begun to focus on the shared couple bed, despite the large proportion of adults who sleep beside their partner (see Troxel, 2010 for a review). This research on couple sleep has highlighted the influence one partner can have on the other partner’s sleep and ultimately their health and well-being (Troxel et al., 2016; Gunn et al., 2016; 2015; Rosenblatt, 2012). The current study examines sleep concordance, specifically the percentage of time partners spend in bed together out of their total time in bed, as a predictor of subjective sleep quality. Research suggests that greater sleep-wake concordance will be associated with greater subjective sleep quality for couples (Richter et al., 2016), however, not everyone may experience these benefits. Attachment security may play a role in how well individuals are able to use their partner for increased feelings of security and stress reduction at bedtime. Therefore, this study investigated attachment style as a moderator of the association between sleep concordance and sleep quality.

Sleep concordance has been defined in multiple ways: whether couples are asleep or awake at the same time during the night (Gunn et al., 2016; 2015), similarity in sleep onset or wake time (Hasler & Troxel, 2010), overlap in physical movements throughout the night (Meadows et al., 2005; Pankhurst & Horne, 1994), and being in the same sleep stage at the same time (Drews et al., 2017). Couple concordance has been observed in all of these ways. For example, couples typically go to sleep and wake up at approximately the same time (Spiegelhalder et al., 2016; Gunn et al., 2016; Gunn et al., 2015). Couples also experience concordance in awakenings and movements throughout the night (Pankhurst & Horne, 1994). However, there has been little research on sleep concordance in time in bed together. Rather than mainly reflecting a form of biological synchrony, concordance in bed time may also represent a conscious choice for partners to be near each other, affectionate, and intimate (Troxel, 2010). To consider this understudied aspect of sleep concordance, the current study defines sleep concordance as the amount of time couples overlap in bed out of the total time at least one partner is in bed.

People may prefer to sleep with their partner because sleep is a vulnerable state that is antithetical to vigilance (Dahl, 1996). Therefore, people need to feel safe in order to reduce vigilance, relax, and fall asleep (Troxel, 2010). According to attachment theory, romantic partners serve as an important source for feelings of safety and security (Hazan & Shaver, 1987), especially during times of vulnerability. Thus, when couples are sleep concordant they may perceive their sleep as subjectively better (Richter et al., 2016) due to added felt security, despite increases in movements and disturbances (Pankhurst & Horne, 1994). Persons with secure attachment style form and keep high quality, close relationships (Bartholomew, 1990). They are able to use their partner as a secure base, from which they gain feelings of safety and protection, buffering against stress from the day, reducing psychological and physiological arousal, making sleep onset easier, and contributing to sounder sleep (Troxel et al., 2016; Troxel, 2010; Troxel et al., 2009; Sbarra & Hazan, 2008). Thus, it is hypothesized that there will be a positive association between sleep concordance and sleep quality for people with greater attachment security.

However, not all couples experience such feelings of security from their partners. Individuals with insecure attachment are unable to use their partner as a secure base, and may not receive the stress-buffering, secure feelings normally associated with sleeping beside a partner (Troxel et al., 2007). For example, persons with avoidant attachment are uncomfortable with intimacy and prefer to be self-reliant rather than seek support from others (Sbarra & Hazan, 2008), suggesting that sleep concordance may have little benefit for them. Persons with anxious-ambivalent attachment experience chronic doubts that partners reciprocate romantic feelings and are trustworthy, view themselves as unworthy of love, and are difficult to comfort and reassure (Sbarra & Hazan, 2008). The implication again is that sleep concordance may offer little benefit for sleep quality. Thus, it is expected that there will be little to no association between sleep concordance and sleep quality among individuals with greater attachment insecurity.

Although this is the first known study to consider attachment security as a moderator of associations between sleep concordance and subjective sleep quality, previous research has conceptualized sleep as an important attachment behavior (Troxel, 2010; Troxel et al., 2007). Additionally, there are mean differences in sleep quality and sleep concordance between individuals with different attachment styles (Gunn et al., 2015). One study using self-report measures of attachment and sleep quality found that attachment anxiety but not avoidance was related to lower sleep quality in both men and women (Carmichael & Reis, 2005). Gunn and colleagues (2015) found that for men, attachment anxiety was associated with greater sleep concordance and that this association was moderated by wives’ marital satisfaction. Thus, there is a strong rationale for expecting associations between sleep concordance and subjective sleep quality to differ based on attachment.

Method

Participants

Data are from a larger study of child and family functioning. Participants were a subset of 179 heterosexual couples from the larger study (N=199) who completed sleep diaries (90% of the total sample). Couples were living together and had a child between the ages of 6 and 12 years. Families were recruited through school systems, after school programs, flyers, mailed post-cards, and family referrals. Study demographics are presented in Table 1.

Table 1:

Study Demographics

	Men	Women
Age, years, M(SD)	40 (7)	38 (7)
Race, n(%)
White	147 (73%)	161 (80%)
Black/African American	28 (14%)	28 (14%)
Hispanic/Latin	4 (2%)	2 (1%)
Asian	5 (2.5%)	2 (1%)
Native American	1 (0.5%)	1 (0.5%)
Other	4 (2%)	3 (1.5%)
Income, n(%)
< $6,000	5 (2.5%)	9 (4.5%)
$6,000 – $11,999	0 (0%)	4 (2%)
$12,000 – $16,999	2 (1%)	2 (1%)
$17,000 – $22,999	4 (2%)	6 (3%)
$23,000 – $28,999	9 (4.5%)	8 (4%)
$29,000 – $39,999	15 (7.5%)	16 (8%)
$40,000 – $54,999	18 (9%)	20 (10%)
$55,000 – $74,999	46 (23%)	44 (22%)
$75,000 – $89,999	24 (12%)	17 (8.5%)
$90,000 – $99,999	11 (5.5%)	19 (9.5%)
$100,000 – $124,999	22 (11%)	20 (10%)
$125,000 – $149,999	13 (6.5%)	14 (7%)
> $150,000	12 (6%)	10 (5%)
Marital Status, n(%)
Married	167 (83%)	172 (85.6%)
Living Together	15 (7.5%)	17 (8.5%)
Other	2 (1%)	7 (3.5%)
Years Living Together, M(SD)	14 (6)	13 (6)

Note: N=179; cells may not add up to 100% due to missing data

Procedure

This study was approved by the University of Kentucky’s Institutional Review Board. To begin the study, a research assistant went to the family’s home in order to obtain informed consent. Then couples were given a form about sleep and functioning to fill out daily for seven days. Couples were called daily to remind them to complete the form. Following the last night of the seven-day assessment, couples came into the laboratory. During part of this three-hour lab visit, couples completed a series of questionnaires.

Measures

Sleep Concordance (measured daily for 7 nights).

Each day couples completed a form indicating their bed time (“What time did you go to bed last night?”), and time out of bed (“What time did you get out of bed?”). Sleep concordance was computed by dividing the total amount of time both partners were in bed together by the total amount of time at least one partner was in bed for each night. Both partners received the same sleep concordance score. The mean level of sleep concordance was 53.64, somewhat lower than previous studies that calculated sleep concordance differently using actigraphy (Gunn et al., 2016; 2015).

Attachment (completed once).

The well-established Spousal Attachment Styles Questionnaire (SASQ; Becker et al., 1997) was used to assess attachment. Participants completed three subscales: fearful (avoidant) attachment (Cronbach’s α=0.82), preoccupied (anxious) attachment (Cronbach’s α=0.88) and secure attachment (Cronbach’s α=0.80) for a total of 25 items rated on a Likert scale from 1 “strongly disagree” to 7 “strongly agree”.

Sleep Quality (completed daily for 7 nights).

Sleep quality was assessed using 4 items, each on a scale from 1 to 10: (1) “What was the quality of your sleep?”, (2) “How alert were you when you first woke up?”, (3) “How rested and refreshed were you when you first woke up?” with 10 being the best, and (4) “How difficult was it for you to get up today?”, with 10 being the most difficult; this item was subsequently reverse scored. Scores were summed for an overall sleep quality score for each day (Cronbach’s α=0.83).

Related Sleep Variables (completed daily for 7 nights).

In addition to bed time and time out of bed, couples indicated their sleep time and wake time. Sleep efficiency was computed by dividing the total amount of time a person was asleep (wake time – sleep time) by the total amount of time each person was in bed (time out of bed – bed time). In addition, two dummy variables were created (0=no, 1=yes) to indicate if the participant went to bed before his or her partner, and if the participant got out of bed after his or her partner.

Additional Covariates.

Caffeine use was measured daily by asking the number of caffeinated beverages consumed. Napping was measured daily by asking if participants took a nap or not. Sickness was measured daily by asking if participants were sick or not. Years living together, income and age were measured once. Drinking problems were assessed once using the Alcohol Use Disorder Identification Test (AUDIT; Saunders et al., 1993). Depression was measured once using the Center for Epidemiological Studies Depression Scale (CESD; Radloff, 1977).

Data Analyses

Data were analyzed using the multilevel model for intensive longitudinal data from distinguishable dyads described by Bolger and Laurenceau (2013). In this approach, models are estimated for male and female sleep quality simultaneously (in the same equation) using dummy variables. The dependent variable was sleep quality. The predictors of interest were sleep concordance, attachment, and the interaction between sleep concordance and attachment. These models also controlled for sleep efficiency and if the participant went to bed before his or her partner (models controlling for if the participant woke up after his or her partner and both if the participant went to bed before his or her partner and if the participant woke up after his or her partner were also fit, but results were identical so only results controlling for if the participant went to bed before his or her partner were presented). Consistent with best practice in multi-level modeling, within-person variables were person-mean centered for level 1 and the person mean was entered at level 2. Separate models were fit for each of the three attachment scales (secure, fearful, and preoccupied). Significant interactions were probed using the online utility for multi-level models developed by Preacher and colleagues (2006). Interactions were plotted to show associations between sleep concordance and sleep quality at +1 and −1 SD from the mean of the moderator. Although the study is correlational and causality cannot be inferred, the term “effect” was used throughout the description and discussion of results to be consistent with how MLM findings are typically described and to simplify presentation.

Preliminary Analyses

Descriptive statistics of study variables are presented in Table 2. Level 1 missing data were handled using full information maximum likelihood (FIML), an approach recommended by methodologists (Wothke, 2000). Based on responses to sleep quality, 1% of men only completed only one or two diary days, 1% of men and 0.5% of women responded to three diary days, 0.5% of men and 1.5% of women completed four diary days, 2% of men and 2.5% of women completed five diary days, 13% of men and 15.5% of women completed six diary days, and 72.5% of men and women completed all seven diary days. Missingness was not associated with any of the predictor variables in the model. There was less than 10% missing data on all level 2 variables included in the final model, thus casewise deletion was used to handle data missing at level 2.

Table 2:

Descriptive Statistics

Measure M(SD)	Men	Women
Sleep Quality	23.45 (4.84)	21.98 (5.12)
Sleep Concordance	53.64 (39.16)	53.64 (39.16)
Sleep Efficiency	0.93 (0.07)	0.93 (0.07)
Bed First	0.38 (0.49)	0.41 (0.49)
Caffeine Use	2.47 (2.14)	1.96 (1.87)
Years Living Together	13.63 (5.66)	13.34 (5.90)
Problem Drinking	3.91 (4.21)	2.73 (2.82)
Secure Attachment	41.98 (7.51)	40.60 (8.71)
Avoidant Attachment	12.15 (7.36)	12.84 (7.99)
Anxious Attachment	16.66 (9.30)	13.57 (7.52)

Note: Bed First refers to whether the participant went to bed before his or her partner (1=yes 0=no).

The effects of covariates were evaluated in two ways, First, by fitting a preliminary model including all potential covariates as predictors of sleep quality (see Table 3). No significant associations were observed. Second, each covariate was separately evaluated in the final model to determine if it had any impact on the significance of other model coefficients. Three covariates, caffeine use, years living together, and problem drinking emerged as potentially important using this second method and are included in the final results.

Table 3:

Associations between Covariates and Sleep Quality

	B		SE		t
	Men	Women	Men	Women	Men	Women
Level 1:
Caffeine Use	0.03	0.02	0.17	0.22	0.14	0.10
Napping	−1.07	−0.22	0.61	0.57	−1.76	−0.39
Sickness	0.82	−0.24	1.10	0.86	0.75	−0.28
Level 2:
Years Living Together	0.03	−0.02	0.08	0.08	0.40	−0.28
Age	0.01	0.02	0.06	0.07	0.18	0.28
Income	−0.01	0.10	0.15	0.13	−0.08	0.73
Problem Drinking	0.05	−0.12	0.11	0.14	0.48	−0.85
Depression	−0.01	−0.001	0.05	0.04	−0.20	−0.04

Note: df=912, No associations are significant at p<.05. Napping refers to whether the participant napped that day or not (1=yes 0=no). Sickness refers to whether the participant was sick that day or not (1=yes 0=no).

Results

Results for moderation by secure, avoidant, and anxious attachment style subscales are presented in tables 4, 5, and 6, respectively. Nights with greater sleep efficiency were related to greater sleep quality for men, Bs=5.86 to 6.27, ps=0.01 to 0.02. Nights with greater caffeine use were marginally related to greater sleep quality for women, B=0.36 to 0.38, ps=0.05 to 0.06. There were some additional marginal associations between sleep efficiency, caffeine use, if the participant went to bed before his or her partner, and daily sleep concordance with subjective sleep quality, but these associations were not consistent across models with secure, anxious, and avoidant attachment.

Table 4:

Sleep Quality as a Function of Sleep Concordance Percentage Moderated by Attachment Security

					CI95
Fixed Effects	Estimate	(SE)	t	pa	Lower	Upper
M_Intercept	9.27	(8.22)	1.13	0.26	−7.05	25.59
F_Intercept	28.22	(8.14)	3.47**	0.001	12.08	44.36
M_Sleep Efficiency	6.27	(2.45)	2.56*	0.01	1.46	11.08
F_Sleep Efficiency	2.92	(3.36)	0.87	0.38	−3.67	9.50
Mean M_Sleep Efficiency	14.78	(8.59)	1.72+	0.09	−2.08	31.64
Mean F_SleepEfficiency	−6.97	(8.77)	−0.80	0.43	−24.18	10.24
M_Bed First	0.27	(0.41)	0.66	0.51	−0.53	1.07
F_Bed First	0.15	(0.47)	0.33	0.74	−0.77	1.07
Mean M_Bed First	2.07	(1.20)	1.72+	0.09	−0.30	4.43
Mean F_Bed First	1.35	(1.09)	1.24	0.22	−0.80	3.50
M_Caffeine Use	−0.03	(0.17)	−0.15	0.88	−0.37	0.32
F_Caffeine Use	0.36	(0.19)	1.88+	0.06	−0.02	0.74
Mean M_Caffeine Use	−0.32	(0.19)	−1.73+	0.08	−0.68	0.04
Mean F_Caffeine Use	−0.29	(0.19)	−1.54	0.12	−0.66	0.08
M_Years Living Together	−0.04	(0.06)	−0.75	0.46	−0.15	0.07
F_Years Living Together	0.07	(0.05)	1.25	0.21	−0.04	0.17
M_Problem Drinking	−0.04	(0.08)	−0.47	0.64	−0.19	0.12
F_Problem Drinking	0.02	(0.12)	0.20	0.86	−0.21	−0.26
M_SC Slope	−0.001	(0.01)	−0.15	0.88	−0.37	0.32
F_SC Slope	0.001	(0.01)	1.88+	0.06	−0.02	0.74
Mean M_SC Slope	0.007	(0.01)	0.55	0.58	−0.02	0.03
Mean F_SC Slope	0.01	(0.01)	0.86	0.39	−0.01	0.03
M_Secure	0.13	(0.10)	1.32	0.19	−0.07	0.33
F_Secure	0.17	(0.07)	2.35*	0.02	0.03	0.32
M_Secure*SC	0.0004	(0.001)	0.65	0.52	−0.001	0.002
F_Secure*SC	−0.001	(0.001)	−0.94	0.35	−0.003	0.001
M_Secure*Mean SC	−0.001	(0.002)	−0.89	0.37	−0.005	0.002
F_Secure*Mean SC	−0.003	(0.001)	−2.31*	0.02	−0.01	−0.001
					CI95b
Random Effects	Estimate	(SE)	z	pa	Lower	Upper
Level-2 (between-couples)c
M_Intercept Variance	9.03	(1.73)	5.21***	<.0001	6.41	13.69
F_Intercept Variance	8.10	(1.73)	4.70***	<.0001	5.56	12.93
M_SC Slope Variance	0d
F_SC Slope Variance	0.001	(0.001)	1.68*	0.05	0.001	0.01
M-F Intercept Covariance	1.91	(1.44)	1.33	0.18	−0.92	4.75
M-F Slope Covariance	0.001	(0.001)	1.26	0.21	−0.0004	0.12
Level-1 (within-couples)
M_Residual Variance	17.97	(1.26)	14.24***	<.0001	15.73	20.73
F_Residual Variance	14.10	(0.98)	14.45***	<.0001	12.37	16.22
M-F Residual Covariance	0.75	(1.15)	0.65	0.52	−1.51	3.00
Autocorrelation	0.06	(0.05)	1.33	0.18	−0.03	0.15

Note: N=179 couples, 7 days. M=male partner, F=female partner SC=sleep concordance. Note: Bed First refers to whether the participant went to bed before his or her partner (1=yes 0=no).

^***p<.001

^**p<.01

^*p<.05

⁺p<.10.

Estimates are unstandardized.

^aAll p-values are two-tailed except for variances with one-tailed p-values.

^bConfidence intervals for variances were computed using the Satterthwaite method (see Littell et al., 2006).

^cCovariances between male intercepts and male SC slopes, female intercepts and female SC slopes, male intercepts and female SC slopes, and female intercepts and male SC slopes were estimated but not included for brevity since none were significant.

^dThe maximum likelihood procedure was unable to generate an estimate for this coefficient as a result of model complexity and therefore it has been fixed to zero.

Table 5:

Sleep Quality as a Function of Sleep Concordance Percentage Moderated by Attachment Avoidance

					CI95
Fixed Effects	Estimate	(SE)	t	pa	Lower	Upper
M_Intercept	7.77	(8.22)	0.95	0.35	−8.54	24.07
F_Intercept	29.16	(8.44)	3.46**	0.001	9.59	41.53
M_Sleep Efficiency	5.88	(2.44)	2.41*	0.02	1.09	10.66
F_Sleep Efficiency	3.05	(3.35)	0.91	0.36	−3.53	9.63
Mean M_Sleep Efficiency	16.32	(8.59)	1.90+	0.06	−0.55	33.19
Mean F_SleepEfficiency	−8.21	(9.07)	−0.90	0.37	−26.03	9.60
M_Bed First	0.28	(0.40)	0.70	0.49	−0.51	1.08
F_Bed First	0.18	(0.47)	0.39	0.70	−0.74	1.10
Mean M_Bed First	2.11	(1.24)	1.69+	0.09	−0.34	4.55
Mean F_Bed First	1.56	(1.10)	1.41	0.16	−0.62	3.74
M_Caffeine Use	−0.03	(0.17)	−0.24	0.81	−0.38	0.30
F_Caffeine Use	0.38	(0.19)	1.95+	0.05	−0.002	0.76
Mean M_Caffeine Use	−0.29	(0.18)	−1.60	0.11	−0.66	0.07
Mean F_Caffeine Use	−0.26	(0.19)	−1.35	0.18	−0.63	0.12
M_Years Living Together	−0.05	(0.06)	−0.84	0.40	−0.16	0.06
F_Years Living Together	0.05	(0.06)	0.92	0.36	−0.06	0.16
M_Problem Drinking	−0.03	(0.08)	−0.32	0.75	−0.18	0.13
F_Problem Drinking	0.03	(0.12)	0.23	0.82	−0.21	0.27
M_SC Slope	−0.002	(0.01)	−0.33	0.74	−0.01	0.01
F_SC Slope	0.0003	(0.01)	0.04	0.97	−0.02	0.02
Mean M_SC Slope	0.01	(0.01)	0.50	0.62	−0.02	0.03
Mean F_SC Slope	0.01	(0.01)	0.94	0.35	−0.01	0.03
M_Avoidant	−0.06	(0.12)	−0.48	0.63	−0.30	0.18
F_Avoidant	−0.18	(0.09)	−2.03*	0.04	−0.35	−0.01
M_Avoidant*SC	−0.0004	(0.001)	−0.68	0.50	−0.002	0.001
F_Avoidant*SC	0.001	(0.001)	1.11	0.27	−0.001	0.002
M_Avoidant*Mean SC	0.00004	(0.002)	0.02	0.98	−0.004	0.004
F_Avoidant*Mean SC	0.003	(0.002)	1.93+	0.05	−0.0001	0.01
					CI95b
Random Effects	Estimate	(SE)	Z	pa	Lower	Upper
Level-2 (between-couples)c
M_Intercept Variance	8.93	(1.71)	5.22***	<.0001	6.34	13.53
F_Intercept Variance	8.11	(1.74)	4.67***	<.0001	5.55	12.97
M_SC Slope Variance	0d
F_SC Slope Variance	0.001	(0.001)	1.68*	0.05	0.001	0.01
M-F Intercept Covariance	2.19	(1.46)	1.50	0.13	−0.67	5.05
M-F Slope Covariance	0.001	(0.001)	1.23	0.22	−0.0004	0.002
Level-1 (within-couples)
M_Residual Variance	18.02	(1.27)	14.22***	<.0001	15.77	20.78
F_Residual Variance	14.37	(0.98)	14.61***	<.0001	12.62	16.51
M-F Residual Covariance	1.08	(1.13)	0.96	0.34	−1.13	3.29
Autocorrelation	0.07	(0.04)	1.54	0.12	−0.02	0.16

Note: N=179 couples, 7 days. M=male partner, F=female partner SC=sleep concordance. Note: Bed First refers to whether the participant went to bed before his or her partner (1=yes 0=no).

^***p<.001

^**p<.01

^*p<.05

⁺p<.10.

Estimates are unstandardized.

^aAll p-values are two-tailed except for variances with one-tailed p-values.

^bConfidence intervals for variances were computed using the Satterthwaite method (see Littell et al., 2006).

^cCovariances between male intercepts and male SC slopes, female intercepts and female SC slopes, male intercepts and female SC slopes, and female intercepts and male SC slopes were estimated but not included for brevity since none were significant.

^dThe maximum likelihood procedure was unable to generate an estimate for this coefficient as a result of model complexity and therefore it has been fixed to zero.

Table 6:

Sleep Quality as a Function of Sleep Concordance Percentage Moderated by Attachment Anxiety

					CI95
Fixed Effects	Estimate	(SE)	t	pa	Lower	Upper
M_Intercept	8.71	(8.18)	1.06	0.29	−7.51	24.93
F_Intercept	24.85	(8.48)	2.93**	0.004	8.03	41.66
M_Sleep Efficiency	5.86	(2.44)	2.41*	0.02	1.09	10.66
F_Sleep Efficiency	3.56	(3.37)	1.06	0.29	−3.06	10.18
Mean M_Sleep Efficiency	−3.43	(9.11)	−0.38	0.71	−21.31	14.45
Mean F_SleepEfficiency	1.78	(1.20)	1.49	0.14	−0.57	4.13
M_Bed First	0.28	(0.41)	0.68	0.50	−0.52	1.07
F_Bed First	0.27	(0.47)	0.58	0.56	−0.65	1.19
Mean M_Bed First	1.78	(1.20)	1.49	0.14	−0.57	4.13
Mean F_Bed First	1.12	(1.13)	0.99	0.32	−1.10	3.35
M_Caffeine Use	−0.03	(0.17)	−0.15	0.88	−0.37	0.32
F_Caffeine Use	0.38	(0.19)	1.95+	0.05	−0.002	0.76
Mean M_Caffeine Use	−0.31	(0.18)	−1.69+	0.09	−0.67	0.05
Mean F_Caffeine Use	−0.23	(0.20)	−1.33	0.18	−0.64	0.12
M_Years Living Together	−0.05	(0.06)	−0.97	0.33	−0.16	0.06
F_Years Living Together	0.06	(0.06)	1.12	0.26	−0.05	0.17
M_Problem Drinking	−0.04	(0.08)	−0.45	0.65	−0.19	0.12
F_Problem Drinking	0.10	(0.13)	0.80	0.42	−0.15	0.36
M_SC Slope	−0.001	(0.01)	−0.24	0.81	−0.01	0.01
F_SC Slope	0.002	(0.01)	0.23	0.82	−0.01	0.02
Mean M_SC Slope	0.005	0.01	0.42	0.68	−0.02	0.03
Mean F_SC Slope	0.01	0.01	0.86	0.39	−0.01	0.04
M_Anxious	−0.06	0.07	−0.80	0.43	−0.20	0.09
F_Anxious	−0.09	0.10	−0.91	0.37	−0.2917	0.1076
M_Anxious*SC	−0.0003	0.001	−0.58	0.56	−0.001	0.001
F_Anxious*SC	0.0004	0.001	0.39	0.69	−0.002	0.002
M_Anxious*Mean SC	0.001	0.001	0.44	0.66	−0.002	0.003
F_Anxious*Mean SC	0.002	0.002	0.90	0.37	−0.002	0.005
					CI95b
Random Effects	Estimate	(SE)	Z	pa	Lower	Upper
Level-2 (between-couples)c
M_Intercept Variance	8.97	(1.74)	5.15***	<.0001	6.34	13.67
F_Intercept Variance	8.79	(1.88)	4.68***	<.0001	6.02	14.04
M_SC Slope Variance	0d
F_SC Slope Variance	0.001	(0.001)	1.70*	0.04	0.001	0.01
M-F Intercept Covariance	2.64	(1.56)	1.70	0.09	−0.41	5.70
M-F Slope Covariance	0.001	(0.001)	1.16	0.25	−0.0004	0.002
Level-1 (within-couples)
M_Residual Variance	18.23	(1.29)	14.09***	<.0001	15.94	21.06
F_Residual Variance	14.50	(1.01)	14.39***	<.0001	12.71	16.70
M-F Residual Covariance	1.17	(1.14)	1.02	0.31	−1.08	3.41
Autocorrelation	0.08	(0.05)	1.71	0.09	−0.01	0.17

Note: N=179 couples, 7 days. M=male partner, F=female partner SC=sleep concordance. Note: Bed First refers to whether the participant went to bed before his or her partner (1=yes 0=no).

^***p<.001

^**p<.01

^*p<.05

⁺p<.10.

Estimates are unstandardized.

^aAll p-values are two-tailed except for variances with one-tailed p-values.

^bConfidence intervals for variances were computed using the Satterthwaite method (see Littell et al., 2006).

^cCovariances between male intercepts and male SC slopes, female intercepts and female SC slopes, male intercepts and female SC slopes, and female intercepts and male SC slopes were estimated but not included for brevity since none were significant.

^dThe maximum likelihood procedure was unable to generate an estimate for this coefficient as a result of model complexity and therefore it has been fixed to zero.

Women who scored higher on the secure attachment subscale (see Table 4) reported greater sleep quality, B=0.17, p=0.02. Additionally, there was an interaction between women’s secure attachment and the person mean of sleep concordance, B=−0.003, p=0.02. Among women with lower scores on the secure attachment subscale (at least 0.73 standard deviations below the mean), there was a positive association between greater person mean of sleep concordance and daily sleep quality; there was no association among women with higher scores on the secure attachment subscale. This interaction is shown in Figure 1A. The interaction has also been plotted with sleep concordance as the moderator and attachment as the independent variable in an attempt to help illustrate significant differences between estimated values of sleep quality based on attachment security. Among women with lower scores on mean sleep concordance (at least 0.82 standard deviations below the mean), there was a positive association between secure attachment and sleep quality. This interaction is shown in Figure 1B.

Figure 1:

Female Interactions with Secure Attachment and Sleep Concordance

Women who scored higher on the avoidant attachment subscale (see Table 5) reported lower sleep quality, B=−0.18, p=0.04. Additionally, there was a marginal interaction between women’s avoidant attachment style and the person mean of sleep concordance, B=0.003, p=0.05. Among women with higher scores on the avoidant attachment subscale (at least 0.92 standard deviations above the mean), there was a positive association between greater person mean of sleep concordance and daily sleep quality; there was no association among women with lower scores on the avoidant attachment subscale. This interaction is shown in Figure 2A. The interaction has also been plotted with sleep concordance as the moderator and attachment as the independent variable in an attempt to help illustrate significant differences between estimated values of sleep quality based on attachment security. Among women with lower scores on mean sleep concordance (at least 1.27 standard deviations below the mean) there was a marginal negative association between avoidant attachment and sleep quality. This interaction is shown in Figure 2B.

Figure 2:

Female Interactions with Avoidant Attachment and Sleep Concordance

There was no association between the anxious attachment style subscale and sleep quality, nor did the anxious attachment style subscale interact with daily sleep concordance or the person mean of sleep concordance in the prediction of sleep quality.

Discussion

The goal of this study was to investigate the association between couple sleep concordance and subjective sleep quality across different levels of attachment security and insecurity. Women who scored higher on the secure attachment style subscale reported higher sleep quality and women who scored higher on the avoidant attachment style subscale reported lower sleep quality. For women with either lower attachment security or higher attachment avoidance, there was a positive association between mean sleep concordance and sleep quality. There were no associations among variables of interest for men.

For women, attachment security may play a role in subjective sleep quality. Women who scored higher on the secure attachment style subscale reported higher sleep quality, whereas women who scored higher on the avoidant attachment style subscale reported lower sleep quality. This adds to previous research that there is a positive association between attachment security and sleep quality (Troxel et al., 2007). Cognitive pre-sleep arousal may be a possible explanation for such findings. One study reported that while trying to fall asleep, “good sleepers” tend to think about nothing in particular, while people with insomnia think more about worries and problems (Harvey, 2000). People with secure attachment style have positive models of the self and others and maintain trusting, intimate, high-quality relationships (Bartholomew, 1990). Perhaps women with greater attachment security do not experience as many troubling thoughts and concerns at bedtime that prevent them from sleeping well (Carmichael & Reis, 2005). Future research may include cognitive pre-sleep arousal as a possible explanation for such associations.

Findings suggest that sleeping beside a partner may be particularly beneficial for the sleep of women who endorse lower levels of attachment security, or alternatively, higher levels of attachment insecurity. This was inconsistent with hypotheses. Perhaps women with greater attachment security are able to rely on themselves and utilize their own coping skills, thus not needing their partner beside them to feel safe and relaxed at night (Carmichael & Reis, 2005). Women with less attachment security may be unable to turn off attachment related worries and self-soothe, experiencing trouble falling asleep, in turn actually benefiting from having their partner beside them in bed for feelings of security and relief (Troxel et al., 2007; Mikulincer & Shaver, 2003).

In the model controlling for secure attachment, there was a marginal positive association between daily sleep concordance and subjective sleep quality for women. This is consistent with previous research that subjective sleep quality is perceived to be better when sleeping in pairs (Spiegelhalder et al., 2016). Spouses often promote health in their partner by monitoring, inhibiting and/or facilitating health behaviors (Waite, 1995). Women in co-sleeping couples may alter their bed time or wake time to match their partners’, contributing to longer total sleep time (Meadows et al., 2009). Additionally, they may fall asleep more easily due to the stress relieving feelings of comfort and security gained by sleeping beside their partner (Troxel, 2010; Troxel et al., 2009). When controlling for attachment avoidance and attachment anxiety, however, this effect was no longer significant and due to its marginal significance should be interpreted carefully.

Overall, these findings suggest that there may be gender differences in the ways sleep concordance and attachment style are associated with sleep quality. There were no significant associations among study variables of interest for men. This is consistent with findings that the association between sleep-wake concordance and health outcomes are stronger for women (Gunn et al., 2016), but are inconsistent with previous associations between anxious attachment and sleep concordance (Gunn et al., 2015). It is possible that sex differences arise from the evolutionary reliance of women on men due to size and dominance for protection from potential predators (Troxel, 2010). Perhaps the beneficial feelings of security women gain by sleeping beside their partner are not shared by men. Additionally, men may not be as impacted by environmental and relational influences as women when trying to fall asleep (Richter et al., 2016).

Findings also support the use of time in bed together in the measurement of sleep concordance. The study of couple sleep concordance is still relatively new, and the multidimensional nature of sleep prevents a single definition from being possible. A variety of measure have been employed, these primarily include objective measures of overlap in sleep-wake states, stages, or physical movements (Drews et al., 2017; Gunn et al., 2016; 2015; Hasler & Troxel, 2010; Meadows et al., 2005; Pankhurst & Horne, 1994). Nevertheless, bedsharing is an important sociocultural aspect of couple sleep concordance, and concordance in time in bed may be an especially important aspect of sleep concordance in relation to attachment.

It is important to interpret these findings in light of study strengths and limitations. This study had a correlational research design; thus, causal inferences cannot be drawn. It is possible that there are bidirectional associations between sleep quality, concordance, and attachment. For example, women who have poor sleep quality may show declines in attachment security over time. Further, multi-level modeling, like other statistical techniques that use maximum likelihood estimation, does not provide reliable standardized estimates to assess the magnitude of associations. In addition, participants in this study were American, heterosexual, predominantly white, middle-class couples with children. Other populations may exhibit a different pattern of results, and future studies with more diverse populations are needed. Industrialized Western societies are unusual when it comes to bed sharing, with the majority of the world sharing their bed with both spouses and children (Mindell et al., 2010). These cultures and cultures where bed sharing is not the norm may shed further light on the examined associations. Additionally, the majority of participants endorsed higher levels of the secure attachment subscale than the insecure attachment subscales. The low amount of variability in attachment style levels may have impacted the significance of results. Finally, future studies should include other variables that may moderate or mediate associations between sleep concordance and sleep quality, such as sexual contact (Dittami et al., 2007), stress and relationship conflict (Troxel et al., 2007), oxytocin (Troxel, 2010), cognitive pre-sleep arousal (Harvey, 2000), and objective measures of sleep quality (Gunn et al., 2015).

Despite limitations, this study had a rigorous seven days of daily reporting on sleep for both partners of heterosexual couples, examined the influence of multiple covariates, and tested an innovative model in which a key moderator of the role of sleep concordance was identified. Findings add to research on dyadic sleep by highlighting the importance sleep concordance may have on sleep quality for women with lower attachment security and higher attachment avoidance. Sleep concordance has benefits beyond affecting sleep quality (Gunn et al., 2015). It promotes intimacy and attachment, giving couples much needed alone time when they may not get it elsewhere, and has important implications for physical and mental health, daily functioning, and well-being (Troxel, 2010). Results suggest this may especially be the case for women.