Bed-Sharing in Couples Is Associated With Increased and Stabilized REM Sleep and Sleep-Stage Synchronization

The study addresses whether sharing a bed with a romantic partner alters sleep neurophysiology and dyadic synchronization of sleep stages. Romantic relationships relate to mental health, and sleep may mediate these effects. Prior actigraphic studies of co-sleep have yielded mixed results on sleep disruption and duration, and only two polysomnographic studies have compared co-sleep to individual sleep in healthy couples, both reporting increased REM sleep but otherwise inconsistent findings. Moreover, the roles of relationship characteristics and chronotype, and the presence of direct dyadic sleep-stage synchronization (including lead–lag effects), remain understudied. The aim was to test if couples sleep better, worse, or differently when co-sleeping and to examine how relational and individual factors influence sleep outcomes and sleep-stage synchrony.

Actigraphic studies indicate bed-sharing may be linked to more disrupted sleep in both sexes or in women only, or to increased sleep time in men, and married couples show longer total sleep time and less wake after sleep onset compared to unmarried singles. Dyadic parameters such as movement synchronization and sleep-wake concordance have been observed, influenced by relationship qualities like conflict or marital quality. However, actigraphy infers sleep from movement and lacks neurophysiological staging. Two polysomnographic studies compared co-sleep to individual sleep in healthy couples: an early lab study reported increased REM sleep and awakenings with fewer changes otherwise; a pilot study by the present authors reported increased REM, SWS, TST, higher sleep efficiency, shorter N2/N3 latencies, and improved subjective sleep. Neither prior study systematically examined relationship characteristics or chronotype as covariates, and only the pilot examined direct sleep-stage synchronization without exploring more complex lead–lag coupling or moderators. Interpersonal synchronization during wake has been linked to prosocial behavior, social bonding, social cognition, and affect, suggesting relevance for sleep-stage synchrony.

Design: Within-subjects crossover in a sleep laboratory. Twelve heterosexual co-sleeping couples (N = 24; ages 18–29) without sleep-affecting disorders, shift work, pregnancy, or relevant medications participated. Inclusion was verified via clinical interview and standard inventories (BDI, AUDIT, CUDIT-R, PSQI, ESS). Procedure: Each couple spent four nights across two consecutive weekends: two nights co-sleeping and two nights individually, with arrangement order counterbalanced. The first night of each weekend served as adaptation and was excluded from analyses. Co-sleep used two adjacent single beds with separate sheets/duvets and a bolstered cleft to create a uniform surface; individual sleep occurred in separate rooms. Couples assigned to individual sleep were separated just before bedtime to standardize pre-sleep activity. Measures: Full cardiorespiratory polysomnography (EEG: F3, F4, C3, C4, O1, O2; EOG; chin EMG; ECG; pulse oximetry; bilateral tibialis anterior EMG; respiratory flow and effort). Sleep staging followed AASM criteria by a blinded rater. Leg movements were automatically detected. REM period defined per sleep cycle; REM fragmentation defined as interruptions between REM epochs within a cycle. Average uninterrupted REM fragment duration was REM duration divided by number of fragments. Dyadic coupling: Cross-recurrence quantification analysis (CRQA) assessed cross-recurrence and leader–follower relationships across time lags, enabling quantification of direct synchrony (lag 0) and lagged synchronization. Subjective sleep: Morning ratings of sleep onset latency, total sleep time, number of awakenings, and a composite morning-condition score (three Likert subscales). Chronotype: German Morningness–Eveningness Questionnaire (D-MEQ). Relationship characteristics: Quality of Relationship Inventory (support, conflict, depth) and Hatfield Passionate Love Scale; relationship duration recorded. Statistical analysis: Confirmatory comparisons of co-sleep vs. individual sleep used paired two-tailed t-tests or Wilcoxon signed-rank tests with normality assessed via Shapiro–Wilk; false discovery rate controlled via Benjamini–Hochberg. Synchronization at lag 0 compared with paired t-tests. Dependence within couples assessed via male–female Pearson correlations to justify individual-level analyses. Exploratory analyses included two-way mixed ANOVAs with within-subject factor sleeping arrangement and between-subject factor gender; ANCOVAs with sleeping arrangement and covariates (conflict, depth, social support, passionate love, relationship duration). Synchronization analyses also included covariates (chronotype, snoring, leg movements). Significance across lags determined by non-overlapping confidence intervals in synchronization plots. Power: Based on expected medium effect size (d = 0.6), α = 0.05, power 0.8, required N = 24 for confirmatory tests. Analyses conducted in R and MATLAB (CRP toolbox); CRQA computed on a high-performance computing cluster.

- Co-sleep vs. individual sleep: REM sleep percentage increased from 21.0 ± 4.2% to 23.0 ± 4.2% of total sleep time (p = 0.005; FDR-corrected p = 0.03), approximately a 10% relative increase. - REM fragmentation decreased during co-sleep (8.5 ± 5.2 vs. 5.4 ± 2.7 interruptions per REM period; p = 0.008; FDR p = 0.03), yielding longer uninterrupted REM fragments (13.4 ± 7.1 vs. 22.0 ± 19.7 minutes; p = 0.0006; FDR p = 0.01). - Limb movements were higher during co-sleep (50.9 ± 32.9 vs. 61.5 ± 36.7; p = 0.007; FDR p = 0.03). - No significant differences in total sleep time, sleep efficiency, sleep onset latency, REM latency, N1, N2, SWS (N3), number of awakenings, apnea–hypopnea index, or snoring events (all ps > 0.085 except awakenings p = 0.148). - Interaction with relationship factors: Social support (QRI subscale) significantly interacted with sleeping arrangement to predict REM percentage (F(1,22) = 4.8, p = 0.039), with larger REM changes in individuals reporting suboptimal support. Other relationship parameters (conflict, depth, passionate love, relationship duration) showed no significant interactions. - Gender: No significant main or interaction effects on REM or other outcomes (all Fs ≤ 0.1, ps ≥ 0.762). - Sleep-stage synchronization: Direct synchrony (lag 0) increased from 36.6 ± 6.0% synchronized epochs during individual sleep to 46.9 ± 8.4% during co-sleep (p = 0.005) when including wake. Excluding wake, synchrony increased from 40.1 ± 7.1% to 47.5 ± 8.9% (p = 0.022). - Lagged coupling: Synchronization during co-sleep exceeded individual sleep across lags, peaking at lag 0. Differences were significant approximately across ±10 minutes when including wake and ±5 minutes when excluding wake. - Relationship depth: Significant positive main effect on synchronization at lag 0 (excluding wake) (F(1,10) = 6.0, p = 0.035). No significant effects of social support, conflict, passionate love, relationship duration, chronotype similarity, snoring, or leg movements on synchrony; sleeping arrangement remained a significant main effect in all tested models (Fs ≥ 6.6, ps ≤ 0.028).

Findings indicate that couples do not simply sleep better or worse when sharing a bed; rather, they sleep differently, with REM sleep specifically increased and stabilized, and with enhanced dyadic synchronization of sleep stages. The REM promotion and reduced fragmentation during co-sleep replicate prior reports and suggest mechanisms beyond mere comfort, potentially including psychosocial safety signals or stress modulation associated with a partner's presence. The interaction with social support implies that relational context modulates REM changes. Increased limb movements during co-sleep, alongside preserved sleep architecture, cautions against interpreting actigraphic movement increases as impaired sleep. Enhanced synchronization, extending across a ±5-minute window and independent of wake periods, suggests complex coupling of partners' sleep architectures. Its positive association with perceived relationship depth links sleep-stage neural synchrony to established effects of interpersonal synchrony during wake on social bonding and cognition. These neurophysiological changes could contribute to a feedback loop wherein social sleeping promotes REM-related functions (emotional memory, stress regulation, social cognition), which in turn may strengthen social functioning and resilience, potentially mediating the protective effects of partnerships on mental health.

Co-sleeping in healthy young couples selectively increases and stabilizes REM sleep and enhances sleep-stage synchronization between partners. Relationship characteristics modulate these effects, with social support interacting with REM changes and relationship depth positively associated with synchrony. These findings support the notion of a feedback loop between sociality and REM sleep and suggest that co-sleep-related neural synchronization may be a mechanism through which relationships support mental health. Future research should test REM-dependent outcomes (memory, affect regulation), examine synchronization in relation to mental health metrics, integrate lab polysomnography with longer-term home actigraphy, employ dyadic statistical models, and include more diverse populations and sleeping arrangements to assess generalizability.

- Laboratory setting may limit ecological validity; results may differ in home environments. - Small sample size for exploratory two-way mixed ANOVAs and ANCOVAs reduces power; negative exploratory findings should be interpreted cautiously. - Multiple testing was not adjusted in exploratory analyses, increasing risk of Type I error. - No dyadic-level statistical modeling; individual-level analysis was used for comparability with prior studies, though dyadic models could yield additional insights. - No assessment of downstream REM-dependent outcomes (memory consolidation, emotional stress dissolution, amygdala reactivity). - Generalizability limited to young, healthy, heterosexual couples; effects in other age groups, clinical populations, and other relationship types remain unknown. - Did not combine polysomnography with extended actigraphy; thus, integration with actigraphic findings in naturalistic settings is limited.