Exposure to war is associated with marked impairments in both sleep and mental health, yet the short-term emotional consequences of war-related nighttime disturbances remain poorly understood. Leveraging nighttime air-raid alerts during the 2025 Israel-Iran war as a naturalistic manipulation, the present study examined their effects on civilians’ sleep and next-day emotions, testing sleep disruption as a mediator and evaluating the roles of siren timing and shelter characteristics.
MethodsA total of 1340 adults residing in Israel (74% female; n = 1140 exposed to nighttime sirens, n = 199 unexposed) were recruited during days 5–11 of the war. Participants completed an online survey assessing nighttime air-raid siren exposure, sheltering conditions, sleep duration, nocturnal wakefulness, and next-day anxiety-depression and mood.
ResultsExposure to nighttime air-raid sirens was associated with 39 fewer minutes of sleep and 34 additional minutes of nocturnal wakefulness compared to non-exposure. Exposed individuals also reported elevated anxiety-depression symptoms and reduced mood the following day. Sleep disruption fully mediated the relationship between nighttime siren exposure and next-day anxiety-depression, and partially mediated effects on mood. Moreover, the magnitude of sleep disruption varied by attack timing and shelter characteristics, with early-morning sirens and external sheltering associated with particularly poor outcomes.
ConclusionsSleep disruption represents a key mechanism linking nighttime attacks to emotional harm in conflict-affected civilians. These findings identify sleep as a critical and modifiable intervention target for protecting mental health during wartime. Interventions to preserve or restore sleep may help mitigate the psychological toll of armed conflict on civilian populations.
Exposure to war has well-documented adverse effects on civilian health and wellbeing (Frounfelker et al., 2018; Kushnir & Melamed, 1992; Landau et al., 1998; Osiichuk & Shepotylo, 2020). A recent World Health Organization report (2025) estimates that roughly one in five individuals exposed to war in the past decade develops a severe mental health disorder. Sleep is among the domains most acutely disrupted, with wartime studies documenting reduced sleep quality and duration, increased insomnia and nightmares, and greater reliance on hypnotics (Mofaz et al., 2023; Rahamim et al., 2025). These disruptions may exacerbate emotional distress and impair daytime functioning, yet the immediate effects of war-related nighttime disturbances on next-day emotions remain largely unexplored. Understanding these short-term impacts is crucial for developing timely interventions during conflict (Haham et al., 2025). Moreover, using real-world war-related nighttime disruptions as a naturalistic manipulation may illuminate fundamental questions about sleep's role in emotion dysregulation. The present study prospectively investigated the effects of nighttime air-raid attacks during the 12-day 2025 Israel-Iran war on civilians' sleep and next-day anxiety-depression and mood, testing sleep disruption as a mediator and examining the roles of attack timing and shelter characteristics.
Research consistently demonstrates the detrimental effects of war on sleep (Cohen et al., 2024; Hoffman et al., 2024; Msaad et al., 2023). Rates of insomnia increase dramatically during wartime; for example, a recent study found a 19 % rise in clinical insomnia from pre-war to wartime in a representative Israeli sample, resulting in a 26 % prevalence during the war (Zak et al., 2025). Similar patterns have been observed in the ongoing Russian-Ukrainian conflict, in which roughly one in five Ukrainian adults (21.8 %) met symptom criteria for insomnia (Khan & Altalbe, 2023). Wartime also brings increased use of hypnotic and psychotropic medications (e.g., benzodiazepines), alongside broader declines in sleep health, duration, and quality relative to pre-war periods (El Haddad et al., 2025; Mofaz et al., 2023; Rahamim et al., 2025; Zak et al., 2025).
Sleep disruption is well established as a contributor to impaired mood and emotional functioning, reflecting the central role of sleep in emotion regulation (Kahn et al., 2013; Palmer & Alfano, 2017). These effects may be enhanced during wartime, when uncertainty and exposure to threat simultaneously disrupt sleep and heighten negative emotions, compounding deficits in emotion regulation (Santiago et al., 2018). A recent study examining the aftermath of a large-scale Iranian missile attack on Israeli civilians in April 2024 found that sleep problems on the night of the attack were associated with elevated distress and anxiety symptoms five days later (Ben-David et al., 2025). These findings suggest that sleep disruption may be a key pathway through which acute war-related events affect wellbeing. However, this study assessed sleep and emotions five days after the attack, raising concerns about recall bias due to retrospective reporting. Moreover, it included only individuals exposed to the attack, precluding a controlled comparison with those who did not experience a missile attack that night.
In June 2025, the Iran-Israel war exposed large civilian populations in Israel to frequent nighttime missile attacks, causing physical damage, casualties, and substantial sleep disruption over a twelve-day period (Human Rights Watch., 2025; United Nations., 2025). Each attack was preceded by a loud air-raid siren alerting civilians to seek shelter, activated selectively in areas predicted to be targeted (e.g., an attack aimed at Tel Aviv would trigger sirens there but not in other regions). Although this occurred within a broader two-year conflict, the Iran-Israel war reintroduced intense consecutive nights of ballistic missile attacks, severely disrupting sleep and heightening feelings of insecurity across the wide civilian population. For example, in the city of Tel Aviv 122 sirens were sounded during the 12-day war, 103 of which occurred during the night (N12, 2025). When sirens sounded, civilians had to quickly reach protective bomb shelters and were asked to remain there for approximately 20 min, substantially disrupting their sleep.
The present study leverages this context to address the methodological limitations noted above, comparing Israeli individuals during an ongoing armed conflict who experienced nighttime air-raid attacks with those who did not on a given night, and examining the effects on same-night sleep and next-day emotions. Assessments were collected on the day immediately following the nighttime event to minimize recall biases and capture the immediate impacts of war-related disruption on sleep duration and quality and next-day anxiety-depression and mood. This design also allowed us to test whether curtailment of sleep or prolonged wakefulness on the night of attacks mediated the impact of attacks on next-day emotions. We hypothesized that experiencing nighttime attacks would be associated with shorter sleep duration and longer nighttime wakefulness, as well as increased next-day anxiety-depression and lower mood. Furthermore, we expected that these disturbances in sleep duration and consolidation would mediate the link between nighttime missile attacks and next-day emotional outcomes.
The study had two additional aims. First, given that attacks occurred at different times during each night throughout the 12 days of war, we examined whether the timing of attacks was associated with the extent of nighttime sleep disruption and next-day anxiety-depression and mood. While nighttime awakenings are known to impair next-day mood and functioning (Finan et al., 2015; Kahn et al., 2014; Medic et al., 2017; Watling et al., 2017), it remains unclear whether the timing of such disruptions is associated with next-day emotional outcomes. Considering that sleep architecture changes across the night, certain time windows of disruption may be more detrimental than others. Disruptions earlier in the night may interrupt slow wave sleep (SWS), which occurs predominantly in the first half of the night and produces more severe sleep inertia upon awakening, impairing performance more than awakenings from other sleep stages (Tassi & Muzet, 2000). Consistent with this, a study found that after a night on-call, physicians’ morning performance on a driving simulator was worse if their sleep had been disturbed in the first third of the night (between 00:00 and 3:00) compared to later in the night (Murray & Dodds, 2003). Conversely, awakenings later in the night are more likely to disrupt REM sleep, which occurs more frequently later in the night and plays a significant role in emotion regulation (Glosemeyer et al., 2020; Pesonen et al., 2024; Walker & van der Helm, 2009). Additionally, resuming sleep after an early morning awakening becomes more difficult once homeostatic sleep pressure has dissipated and the circadian drive for alertness increases (Borbély, 2022). Given the sporadic nature of attacks across different nighttime hours during this 12-day war, we were able to examine timing effects exploratorily, without a specific hypothesis regarding which disruption window would be most detrimental.
Second, we sought to test whether different sheltering conditions were associated with sleep quality and duration, as well as next-day emotions. Each air-raid siren signaled an incoming missile attack and prompted civilians to seek shelter- a system that saves lives and is widely used during wartime worldwide (Van Dijcke et al., 2023). During the current conflict, some civilians were able to remain in fortified in-home shelters, whereas others had to seek protection in shared building shelters or public shelters. For those with in-home shelters, the ability to sleep inside the shelter may mitigate sleep disruption during nighttime attacks. We therefore assessed whether shelter type was associated with nighttime sleep and next-day emotions. Furthermore, we considered the social context of sheltering. Prior research indicates that community resilience and a strong sense of coherence can buffer stress responses during crises (Braun-Lewensohn & Sagy, 2014), raising the possibility that being with others (e.g., family members, friends, or neighbors) during an attack could be linked with the extent of sleep disruption and next-day emotions. We thus hypothesized that in-home sheltering and being with others rather than alone would be associated with better nighttime sleep and next-day emotional outcomes.
MethodParticipants and procedureThe sample included 1340 adults aged ≥18 (74 % female). Individuals were invited to participate in an online survey through online advertisements on social media platforms (Instagram, Facebook, and Whatsapp), as well as via word-of-mouth. Recruitment was ongoing during days 5–11 of the Iran-Israel war (June 17–23, 2025), and the online survey was open for submissions throughout this period. A ceasefire came into effect on day 12 (June 24, 2025). During the recruitment period, sirens sounded both during the day and at night in many areas of Israel, calling citizens to seek shelter from missile attacks. Inclusion criteria were: (1) age 18 or older; and (2) residing in Israel during the war. Demographic information of participants is presented in Table 1.
Demographic and sleep characteristics of the entire sample and stratified by exposure to sirens on the previous night.a.
Data collection was anonymous, and the study was approved by the Institutional Review Board of Tel Aviv University, #0010,662–1. After providing informed consent electronically, participants completed a 5–10-minute online survey (median completion time was 4.58 min) via the Qualtrics platform. The survey included items assessing whether and when participants experienced a nighttime air-raid siren the previous night, how they sheltered, their sleep duration and nocturnal wakefulness, and their current day emotional state. Demographic characteristics were also collected. As compensation, participants who completed the survey were entered into a raffle to receive 10 monetary prizes of approximately 30 USD each.
A total of 1547 survey entries were initially recorded. Only participants who completed the survey prior to the ceasefire on day 12 of the war were retained, resulting in 1460 respondents. Of these, 26 participants did not provide any survey responses beyond initial access, and 30 participants provided only gender and/or age information, and were thus excluded. An additional 64 participants discontinued the survey prior to the sleep module and therefore did not provide any data regarding sleep or next-day emotions. Because these measures constituted the primary outcomes of interest, these participants were excluded, and the final analytic sample comprised 1340 participants. To examine whether participants who discontinued prior to the sleep module differed from those included in the analytic sample, we compared the partial completers to the 1340 included participants on available baseline variables. No significant differences were observed in gender, age group, or shelter type (all ps > 0.05). However, a small but statistically significant association was observed for nighttime siren exposure (χ²(1) = 17.56, p < 0.001, φ = 0.11), indicating that participants who discontinued were modestly less likely to report experiencing a nighttime siren. Additional attrition occurred in later questionnaire sections (e.g., mood and demographic variables), resulting in varying sample sizes across specific analyses (Ns ranging from 1239 to 1340, as reflected in Table 1). All models were estimated using available data for the variables included in each analysis.
MeasuresAir Raid Siren Occurrence and Timing: Four variables were used to assess siren occurrence and timing. Participants reported (1) whether they experienced any sirens during the night, and (2) the timing of nighttime siren occurrences by selecting from predefined two-hour intervals (20:00–22:00, 22:00–00:00, 00:00–02:00, 02:00–04:00, 04:00–06:00, 06:00–08:00). Many participants experienced multiple sirens during the previous night, and were therefore able to select multiple time windows. The total number of siren intervals selected was recorded to create a variable labeled (3) ‘Number of Nighttime Siren Windows’. This was used as a proxy for the extent of nighttime disruption caused by sirens. (4) Participants were also asked whether they experienced additional daytime sirens since waking up.
Shelter Type, Company, and Location: Participants provided information about how they protected themselves during nighttime sirens. Shelter type was categorized as: (1) in-home fortified shelter (a reinforced protected space inside an apartment or house, allowing individuals to reach it quickly), (2) shared building shelter (located in the shared floor or basement, accessible to all apartment building residents), or (3) public shelter (located outside the house or apartment building, usually larger and shared by more residents). Additional variables included: who accompanied the participant in the shelter (alone, family, friends, and/or neighbors), whether they slept in the protected shelter overnight, whether they spent the night at their own home or elsewhere (as some individuals relocated to more protected areas during the war).
Sleep Metrics: Participants were asked about their sleep during the previous night, including sleep onset time [‘What time did you go to sleep last night?’], Wake After Sleep Onset [WASO; ‘How long were you awake during the night?’], and Total Sleep Time [TST; ‘How many hours did you sleep in total last night, not including the time you were awake during the night?’]. These items were derived from the consensus sleep diary (Carney et al., 2012). Consistent with contemporary practice, the consensus-sleep diary was delivered online (Dietch & Taylor, 2021). Participants were also asked to report their habitual TST under regular circumstances, outside wartime.
Anxiety and Depression: The brief screening scale of the Patient Health Questionnaire-4 (PHQ-4; Kroenke et al., 2009) was used to measure depression and anxiety symptoms on the day directly following the assessed night. The PHQ-4 is commonly used to briefly assess anxiety and depression symptoms (Bendau et al., 2021; Jaradat et al., 2020). It consists of 4 items, rated on a 4-point Likert scale. These include the Generalized Anxiety Disorder 2-item scale (GAD-2, “Feeling nervous, anxious, or on edge” and “Not being able to stop or control worrying”; Kroenke et al., 2007), as well as the Patient Health Questionnaire 2-item scale (PHQ-2, “Little interest or pleasure in doing things” and “Feeling down, depressed, or hopeless”; Kroenke et al., 2003). Instructions were adapted for daily use to capture symptoms experienced on the day of assessment (Bautista et al., 2023; Patrick et al., 2022). The four items were averaged to create an anxiety-depression score, and demonstrated good internal reliability (α = 0.80).
Mood: Participants rated their overall mood on the assessment day by responding to the question “How did you feel today in general?” on an 11-point scale ranging from 0 (very bad) to 10 (very good). Single-item measures using 0–10 scales are a well-established approach for assessing current day mood (e.g., de Bloom et al., 2013; Horwitz et al., 2022).
Demographics: Information was collected regarding participants’ age, gender, familial status, occupational status, education level, socioeconomic status, cultural background, the presence of any diagnosed sleep disorders, and location during the assessed night. Additionally, the city in which participants resided during the night of measurement was classified based on attack frequency during the war, using data from the Israel Central Bureau of Statistics. This created an objective measure of city siren exposure level.
Data analysis planStatistical analyses were conducted using JASP (version 0.19.1; JASP Team, 2023) and R (laavan package; Rosseel, 2012). WASO, TST, habitual TST, and demographic characteristics were first compared between those who experienced a nighttime siren to those who didn’t using t-tests for continuous variables and χ² tests for categorical variables (See Table 1). For categorical variables with more than two levels, post-hoc comparisons were performed using the Standardized Residual Method. In addition, a generalized linear mixed model (GLMM) was conducted to compare between TST during wartime to habitual TST on non-war nights. This model included a random intercept for each participant to account for within-subject dependency and controlled for relevant demographic and contextual covariates identified in the preliminary analyses. Effect sizes were reported to quantify the magnitude of differences (φ or Cramér’s V for chi-square tests; Cohen’s d for t-tests). In all following analyses, effect sizes for statistically significant effects were reported as Cohen’s d for 2-level comparisons or standardized regression coefficients (β) for continuous predictors.
Generalized linear models (GLMs) were used to assess associations between siren exposure and nighttime sleep variables (TST and WASO), next-day variables (anxiety-depression and mood), and shelter variables (type and company). Models were adjusted for the following covariates, using backwards selection procedures (Sauer et al., 2013): age, gender, socioeconomic status, parental status, city exposure level, and daytime sirens. Additionally, habitual TST was adjusted for when modelling TST (on the previous night). See supplementary 1 for the included covariates of each analysis. Significant main effects for categorical variables with >2 levels were interpreted using post hoc pairwise comparisons with Tukey’s Honest Significant Difference (HSD) tests.
For variables in which participants could select multiple options (i.e., number of nighttime siren windows and shelter company), responses were dummy-coded. Each option was represented as a separate binary variable (0 = absent, 1 = present). For the variable shelter company, separate dummy variables were created for the presence of family, friends, and neighbors, with “alone” serving as the reference category. This allowed participants to be coded as “1″ on multiple categories simultaneously (e.g., with both family and neighbors in the shelter). Similarly, each two-hour siren interval (20:00–22:00, 22:00–00:00, 00:00–02:00, 02:00–04:00, 04:00–06:00) was coded as a distinct binary variable, with “no siren” as the reference category. Each dummy variable was entered separately into the GLM as an independent predictor.
Finally, to examine whether sleep metrics mediated the relationship between nighttime siren exposure and next-day anxiety-depression and mood, mediation analyses were conducted using the lavaan package in R. Nighttime siren exposure (yes/no) was entered as the independent variable, TST and WASO served as mediators, and next-day anxiety-depression symptoms and mood served as dependent variables. Prior to model estimation, multivariate normality was assessed using Mardia’s test, which indicated significant deviations from normality. Accordingly, all mediation models were estimated using robust maximum likelihood (MLR), which provides standard errors and test statistics that are robust to non-normality. Indirect (a × b), direct (c′), and total (c) effects were estimated, and standardized coefficients were reported.
ResultsPreliminary analysesParticipants who experienced nighttime sirens on the night of measurement showed no significant differences from those who did not on most demographic characteristics, except for SES (χ²(3) = 9.671, p = 0.02, V = 0.09) and sector (χ²(4) = 16.75, p = 0.002, V = 0.12; See Table 1). Participants who experienced nighttime sirens were more likely to be Secular Jewish (p < 0.01) and less likely to be Religious/Orthodox Jewish (p < 0.01) compared to those who did not have nighttime siren exposure. Additionally, participants who were exposed to a nighttime siren were more likely to be from middle-high SES (p = 0.02) and less likely to be from low-middle SES (p = 0.02) compared to those who did not experience a nighttime siren. No other Sector or SES group differences were statistically significant. These mentioned differences between groups in siren exposure may reflect differences in the populations residing in cities that were more heavily targeted than others.
Furthermore, in terms of sleep duration, GLMM revealed a significant main effect of wartime. Participants reported sleeping less during wartime compared to habitual nights regardless of siren exposure (χ²(1) =676.94, p < 0.001, Cohen’s d = 1.19). Estimated marginal means indicated an average TST of 7.38 h (SE=0.04) during habitual non-war nights and 6.16 h (SE=0.04) during wartime nights, reflecting a difference of approximately 73 min of sleep on average.
Is nighttime air-raid siren exposure associated with sleep and emotions?GLMs were used to examine whether the occurrence of nighttime sirens during wartime was associated with changes in TST and WASO, as well as next day anxiety-depression symptoms and mood. Results revealed that participants who experienced a nighttime siren slept approximately 39 min less (SE = 0.10, t = –6.32, p < 0.001, Cohen’s d = 0.49) and were awake for approximately 34 min longer during the night (SE = 3.39, t = 10.08, p < 0.001, Cohen’s d = 0.78) compared to those who did not experience a nighttime siren. These findings indicate a negative effect of nighttime sirens on both sleep duration and continuity. Additionally, experiencing a nighttime siren was associated with higher next-day anxiety-depression scores (Estimate = 0.13, SE = 0.06, t = 2.13, p = 0.03, Cohen’s d = 0.17) and lower mood (Estimate = –0.72, SE = 0.16, t = –4.44, p < 0.001, Cohen’s d = 0.37).
Moreover, a greater number of disrupted nighttime windows was associated with both shorter sleep duration (SE = 0.04, t = –9.61, p < 0.001, β = −0.25) and longer nocturnal wakefulness (SE = 1.43, t = 12.16, p < 0.001, β = 0.31). Specifically, each additional window of reported siren occurrence was associated with approximately 25 min less TST and 17 min more WASO on average. A higher number of nighttime windows in which participants reported sirens occurring was also associated with increased anxiety-depression symptoms (Estimate = 0.08, SE = 0.03, t = 3.08, p = 0.002, β = 0.09) and lower mood (Estimate = –0.36, SE = 0.07, t = –5.24, p < 0.001, β = −0.15).
Does sleep disruption mediate the effects of nighttime sirens on next-day emotions?Mediation analysis revealed that the association between nighttime siren occurrence and next-day anxiety-depression scores was fully mediated by sleep duration and continuity during the night of exposure (Fig. 1). Participants who reported experiencing nighttime sirens had shorter TST (a = –0.69, 95 % CI [−0.896, −0.477], p < 0.001) and longer WASO (a = 34.53, 95 % CI [28.023, 41.045], p < 0.001), and in turn, these sleep disruptions were associated with higher next-day anxiety-depression (for TST: b = –0.11, 95 % CI [−0.139, −0.078], p < 0.001; for WASO: b = 0.004, 95 % CI [0.003, 0.005], p < 0.001). Indirect effects were significant for both mediators (TST: c = 0.08, 95 % CI [0.04, 0.11], p < 0.001; WASO: c = 0.13, 95 % CI [0.09, 0.17], p < 0.001), whereas direct effects were nonsignificant (TST: c′ = 0.02, 95 % CI [−0.093, 0.130], p = 0.75; WASO: c′ = –0.04, 95 % CI [−0.156, 0.076], p = 0.50). These findings indicate that reduced sleep duration and increased nocturnal wakefulness serve as the key pathways linking nighttime siren exposure to poorer next-day emotional well-being.
Conceptual mediation model examining the indirect effect of Nighttime Siren Exposure on next-day emotional outcomes through sleep. Nighttime Siren Exposure (X) was modeled as the predictor. Total Sleep Time (TST) or Wake After Sleep Onset (WASO) served as the mediator (M) in separate models. Next-day anxiety-depression scores or next-day mood scores were modeled as the outcomes (Y).
Mediation modeling similarly revealed that sleep duration partially mediated the association between nighttime siren exposure and next-day mood. Siren exposure was associated with shorter TST (a = –0.69, 95 % CI [–0.91, –0.47], p < 0.001), and shorter TST was associated with lower next-day mood (b = 0.35, 95 % CI [0.27, 0.43], p < 0.001). The indirect effect through TST was significant (c = –0.24, 95 % CI [–0.34, –0.15], p < 0.001), while the direct effect remained significant (c′ = –0.41, 95 % CI [–0.71, –0.12], p = 0.006), indicating partial mediation. Additionally, prolonged nocturnal wakefulness fully mediated this association. Participants exposed to nighttime sirens experienced significantly longer WASO (a = 35.23, 95 % CI [28.51, 41.94], p < 0.001), which in turn predicted lower next-day mood (b = –0.012, 95 % CI [–0.014, –0.009], p < 0.001). The indirect effect through WASO was significant (c = –0.41, 95 % CI [–0.53, –0.29], p < 0.001), and the direct effect was no longer significant (c′ = –0.22, 95 % CI [–0.52, 0.08], p = 0.16), consistent with full mediation.
Is the timing of sirens associated with sleep and next-day emotions?Generalized linear models (GLMs) examined whether the timing of nighttime air raid sirens was associated with sleep disruption and next-day emotions. Six time windows (20:00–22:00, 22:00–00:00, 00:00–02:00, 02:00–04:00, 04:00–06:00, and 06:00–08:00) were entered simultaneously, with participants who did not experience any nighttime sirens serving as the reference group (See Fig. 2). Overall, these models demonstrated that the timing of sirens was significantly associated with sleep disruption and next-day emotions. Across outcomes, several time windows differed significantly from the no-siren reference group, with siren exposure generally associated with shorter TST, longer WASO, worse mood, and higher anxiety-depression symptoms.
The effect of nighttime siren timing on sleep and next-day emotional outcomes. This figure presents results from four separate Generalized Linear Models examining the impact of siren timing across different 2-hour time windows (Y-axes) on four outcomes (X-axes): Total Sleep Time (TST) in hours (A), Wake After Sleep Onset (WASO) in minutes (B), anxiety-depression symptoms (C), and Mood (D). The dashed orange vertical line in each panel represents the mean score of the reference group (participants who did not experience a nighttime siren). For each time window, the circle shows the estimated mean and the horizontal line represents the standard error. Comparisons where the mean is significantly different from the reference group (p < 0.05) are indicated by green, while grey signifies non-significant differences.
When examining TST, sirens occurring in nearly all time windows were associated with significantly shorter sleep durations compared to the no-siren mean (∼6.6 h; see Fig. 2A). For example, sirens occurring between 06:00–08:00 were associated with a 32-minute shorter sleep duration compared to the no-siren reference (SE = 0.12, t = −4.47, p < 0.001, Cohen’s d = 0.41). The only exception was the 22:00–00:00 window, where TST did not differ from the no-siren group (Estimate = −0.11, SE = 0.10, t = −1.03, p = 0.31). Sleep fragmentation patterns were also associated with siren timing (Fig. 2B). WASO was significantly longer for sirens occurring in the early evening (20:00–22:00) and the post-midnight block (00:00–06:00; see Supplementary Table 2 for full model results and all comparison estimates). Descriptively, the 02:00–04:00 window appeared most strongly associated with sleep continuity, showing a 23-minute increase in WASO relative to individuals who did not experience a nighttime siren (SE = 2.63, t = 8.91, p < 0.001, Cohen’s d = 0.54). Experiencing a siren between 06:00–08:00, however, did not differ significantly from not experiencing a nighttime attack in terms of WASO (Estimate = 2.53, SE = 3.95, t = 0.64, p = 0.52), likely reflecting sleep termination following the attack.
Next-day emotional outcomes showed distinct time effects. For anxiety-depression symptoms (Fig. 2C), significant increases over the reference baseline were observed specifically following sirens at 00:00–02:00 and 06:00–08:00 (see Supplementary Table 2 for all comparison estimates). Next-day mood appeared more sensitive to disruption, with participants reporting significantly lower mood following sirens in the early evening (20:00–22:00), across the 00:00–04:00 block, and in the early morning (06:00–08:00; See Fig. 2D and Supplementary Table 2). Notably, the 22:00–00:00 window was the only time period consistently unassociated with negative outcomes across all four sleep and emotional measures.
Is shelter type associated with sleep and next-day emotions?Modelling revealed a significant effect of shelter type on both TST (t = −3.20, p = 0.001) and WASO (t = 4.63, p < 0.001; See Table 2). Post-hoc comparisons found that participants who sheltered inside their homes slept significantly longer and were awake for shorter periods of time during the night than those who sheltered in shared building shelters. Similarly, they slept longer and had shorter WASO than participants who sheltered in public shelters outside their building. The two external shelter types did not differ significantly from each other in either TST or WASO. With regards to next-day anxiety-depression symptoms and mood, no significant effects for shelter type were observed.
Nighttime sleep duration, wake after sleep onset, next-day anxiety-depression, and next-day mood scores by shelter types.a.
Next, we examined whether different types of company during the nighttime attack (family, friends, or neighbors) predicted sleep and emotional outcomes compared to having no company. Because participants could endorse more than one category (e.g., sheltering with both friends, family, and neighbors), the three company types were dummy coded and entered concurrently into GLMs, with “being alone” serving as the reference group. Modelling of TST revealed that the type of company was not associated with nighttime sleep duration. Similarly, no consistent effects emerged for WASO, with one exception: participants who sheltered with neighbors were awake approximately 8 min longer during the night (SE = 3.08, t = 2.62, p = 0.009, Cohen’s d = 0.19) than those who were alone. Regarding emotional outcomes, being with family, friends, and/or neighbors versus being alone was not associated with anxiety-depression symptoms or mood. Overall, and contrary to expectations, the type of social company during the attack was largely unrelated to either sleep or next-day emotional outcomes.
DiscussionThe present study examined the effects of nighttime air-raid sirens during the Israel-Iran 12-day war on civilian sleep and next-day emotions. Using a naturalistic design within a large sample, we found that nighttime sirens were robustly associated with both curtailed sleep duration and increased nocturnal wakefulness, as well as with next-day higher anxiety-depression symptoms and lower mood. Importantly, mediation analyses demonstrated that effects on next-day emotions were largely explained by same-night sleep disruption, highlighting sleep loss and fragmentation as key proximal mechanisms linking war-related nighttime events to next-day emotional distress.
Compared to individuals who did not experience a nighttime air-raid siren, those who did slept approximately 39 min less and were awake 34 min longer during the night. These effects were dose-dependent, with each additional two-hour window of nighttime siren exposure associated with approximately 25 min less sleep and 17 min more nighttime wakefulness. Beyond the acute effects of nighttime sirens, participants overall showed substantial sleep curtailment compared to their habitual pre-war sleep patterns, sleeping approximately 73 min less per night during wartime regardless of whether they experienced a nighttime siren on the assessed night. These findings extend previous research documenting widespread sleep disruption during conflict (Cohen et al., 2024; El Haddad et al., 2025; Hoffman et al., 2024; Msaad et al., 2023; Zak et al., 2025). Furthermore, they underscore the particular vulnerability of sleep to acute nighttime disruptions. Beyond generalized wartime stress, sirens force abrupt awakenings, rapid mobilization, and prolonged hyperarousal while sheltering, all of which directly interfere with sleep continuity and the ability to return to sleep (Horner, 1996). The present results provide rare quantitative estimates of this disruption in a large civilian sample assessed in real time during active conflict.
A central finding of this study was that sleep disruption serves as a major pathway linking nighttime siren exposure and next-day emotions. Participants who experienced nighttime sirens reported higher anxiety-depression scores and lower mood on the following day compared to those who did not experience a nighttime attack. Mediation analysis revealed that both reduced sleep duration and increased nocturnal wakefulness fully mediated the relationship between exposure to nighttime sirens and next-day anxiety-depression symptoms. For next-day mood, nocturnal wakefulness fully mediated and sleep duration partially mediated the relationship with nighttime siren exposure. These findings align with previous literature showing that exposure to threat during wartime both disrupts sleep and increases negative emotions (Santiago et al., 2018), and extend these findings by suggesting that sleep disruption is a key mechanism through which nighttime attacks affect emotional wellbeing. Our results are consistent with theoretical models emphasizing sleep's critical role in emotion regulation, facilitating the processing of emotional experiences and supporting the regulation of affective responses (Gruber & Cassoff, 2014; Kahn et al., 2013; Walker & van der Helm, 2009). When sleep is disrupted, these regulatory processes are impaired, leaving individuals more vulnerable to emotional distress. During wartime, when both sleep disruption and emotional demands are heightened, this vulnerability may be especially pronounced.
Our analyses further revealed that the timing of air-raid sirens was differentially associated with sleep and next-day emotional outcomes. Notably, the 22:00–00:00 window emerged as a period of relative resilience; sirens during this interval were not significantly associated with sleep disruption or next-day emotions. This pattern may reflect that 22:00–00:00 represents a transitional 'downtime' period after familial and professional obligations but before sleep onset, when disruptions are less consequential. Additionally, delayed sleep onset during the war, driven by anticipation of nighttime sirens and facilitated by school closures and remote or suspended work schedules, may have meant many individuals had not yet fallen asleep during this window. In contrast, sirens occurring later in the night (after 00:00), and particularly in the early morning (06:00–08:00), were largely associated with poorer sleep and emotional outcomes. The 06:00–08:00 window was characterized by marked reductions in sleep duration without increased fragmentation, consistent with a reduced ability to return to sleep as homeostatic sleep pressure dissipates and circadian-driven alertness increases near habitual wake times (Borbély, 2022). Early-morning sirens may therefore effectively terminate the sleep episode and initiate the day under conditions of unrecovered sleep loss and acute threat, thereby amplifying next-day anxiety-depression symptoms and lowering mood. Additionally, disruptions during REM-rich portions of the night, which predominate in the latter part of the sleep period, may disproportionately affect emotional regulation processes (Glosemeyer et al., 2020; Walker & van der Helm, 2009). This interpretation aligns with evidence from insomnia research showing that profiles characterized by early-morning awakenings are associated with greater depressive symptomatology than sleep-maintenance difficulties alone (Bragantini et al., 2020). Although these explanations remain speculative, our findings highlight the importance of temporal characteristics of sleep disruption, suggesting that the emotional impact of nighttime stressors depends not only on whether sleep is disturbed, but also on when the disruption occurs and whether recovery remains possible.
We additionally sought to examine how the means of protection during an attack influenced sleep and emotional outcomes. Shelter type was significantly associated with nighttime sleep, with individuals using in-home fortified rooms exhibiting better sleep than those using shared building or public street shelters. Compared with in-home shelters, use of shared building shelters was associated with approximately 8 min less total sleep time and 6 additional minutes of wake after sleep onset, while use of public shelters was associated with reductions of 16 min in total sleep time and 13 additional minutes of wakefulness. These findings underscore the relative protective value of in-home shelters for preserving sleep during nighttime sirens. Remaining in a familiar and private environment likely reduces both physical and psychological disruption. However, shelter type was not associated with next-day mood and anxiety-depression symptoms. This finding is in line with prior work showing that being near a shelter during a missile attack was not associated with peritraumatic distress and generalized anxiety five days after the attack (Ben-David et al., 2025). One explanation is that attacks create a sense of unsafety that extends across shelter types. Additionally, reports of civilian casualties inside in-home shelters during this conflict (Jerusalem Post, 2025), which had been rare in the past, may have increased feelings of vulnerability even in perceived safe spaces, leading to similar mood and anxiety-depression symptoms across shelter types. The observed sleep differences, while statistically significant, may have been too modest to produce differential emotional effects, with wartime emotional responses being shaped by factors beyond shelter characteristics.
We also investigated the social context of sheltering, considering whether the presence of family, friends, neighbors, or being alone was associated with sleep and well-being. Contrary to expectations, social company during attacks was not consistently associated with sleep or next-day emotional outcomes. The only exception was that sheltering with neighbors was associated with slightly longer nocturnal wakefulness, possibly reflecting crowding, noise, or discomfort in public shelters. The lack of clear protective effects of social company was surprising given research documenting the buffering role of social support and community resilience during crises (Braun-Lewensohn & Sagy, 2014). It is possible that the context of sheltering (e.g., crowded conditions, heightened anxiety) may have inhibited supportive social interactions from occurring or even introduced additional sources of arousal that counteract potential buffering effects. Another possibility is that social support operates through perceived support rather than physical presence during the event itself. Our study did not assess perceived social support, which may be more strongly linked to emotional well-being than actual company during attacks. Indeed, perceived social support has been associated with lower stress, higher positive affect, and reduced depression and anxiety, including during periods of acute stress (Acoba, 2024; Chin et al., 2023; Wang et al., 2018). While some participants reported being physically alone during the nightly attack, they may have received support from their loved ones throughout the day, or via phone calls and messages. Future studies could assess these dimensions more directly to better characterize social protective factors during wartime.
Strengths and limitationsThis study has several notable strengths. The naturalistic design leveraged real-world variation in nighttime attack exposure, providing ecological validity while allowing for controlled comparisons between exposed and unexposed individuals. Data were collected during the active conflict period, minimizing recall bias and capturing the immediate impacts of attacks on sleep and emotions. The large sample size enabled detection of relatively small effects and examination of moderating factors such as siren timing and shelter characteristics. Finally, the mediation analyses provided insight into the mechanisms linking nighttime attack and siren exposure to emotional outcomes, identifying sleep disruption as a major pathway.
Despite these strengths, several limitations should be acknowledged. First, the study relied on self-reported sleep measures, as objective assessments (e.g., actigraphy) were impractical during a period of high-intensity conflict. Additionally, although sleep and mood during the wartime period were assessed in close temporal proximity to the nighttime siren exposure, habitual sleep prior to the war was assessed retrospectively. Therefore, reports of habitual sleep may be subject to recall bias, potentially affecting comparisons between wartime and pre-war sleep patterns. Moreover, sleep, mood, and anxiety-depression symptoms were assessed using brief or single-item measures to reduce participant burden and maximize compliance under stressful conditions. Future studies could incorporate objective sleep measures to provide more precise estimates and reduce recall bias (Sadeh, 2015). Second, the sample was a convenience sample, which may not be fully representative of the broader Israeli population or of civilians in other cultural or geopolitical contexts. Participants were drawn from a population with prior exposure to missile attacks, which may limit generalizability to less conflict-exposed settings. Additionally, although no significant demographic differences were found for most variables between the siren-exposed and non-exposed groups, differences emerged in SES and sector. Considering that attacks were geographically targeted, the most affected areas often included central urban regions containing strategic landmarks (e.g., hospitals, universities). For example, Tel Aviv is a densely populated metropolitan area that was frequently targeted during the war, and is also characterized by relatively higher average SES and a larger proportion of secular residents. Thus, the observed differences in SES and sector may reflect residential distributions rather than systematic sampling bias, whereby individuals living in central, higher-SES, and more secular municipalities were more likely to be exposed to nighttime sirens. Importantly, the sample was large, socioeconomically diverse, and included various age groups and education levels.
Third, attrition occurred across survey sections. Approximately 6 % (64) of initial survey responders discontinued prior to the sleep module, and further attrition (up to 15 %) occurred for later survey sections. Missingness was not completely random: participants who discontinued were modestly less likely to report experiencing a nighttime siren. However, no other differences were observed between included and excluded participants, and the magnitude of the exposure-related attrition effect was small. Nevertheless, selective attrition may have introduced a minor bias and should be considered when interpreting the findings. Fourth, participants did not report the exact number of nighttime sirens; instead, exposure was indexed by the number of nighttime windows in which sirens occurred. Although comparisons between nights with and without sirens were informative, the lack of a non-war control group limits our ability to disentangle war-specific effects from more general stress-related processes. In addition, while sleep occurred prior to next-day emotions, the observational nature of the data precludes causal inference. Experimental manipulation of sleep in wartime contexts is neither feasible nor ethical; however, future research could employ ecological momentary assessment or intensive longitudinal designs to better capture within-person temporal dynamics. Finally, each participant completed only a single assessment, limiting our ability to track changes across different stages of the conflict or into the post-war period. Future studies conducted during conflict should aim to incorporate repeated assessments of each participant both during the conflict in its aftermath.
Clinical implications and conclusionsThis study highlights the immediate toll that nighttime air-raid sirens during active war take on civilians’ sleep and wellbeing, with sleep loss and fragmentation emerging as central and modifiable mechanisms through which acute wartime stressors shape next-day emotions. The timing of sirens and access to in-home shelters moderated these effects, with early morning and mid-night sirens, as well as external sheltering associated with particularly poor outcomes. These findings underscore sleep as a critical vulnerability and potential intervention target during wartime. Protecting civilian sleep through infrastructure investments, such as widespread access to in-home shelters, may benefit both physical safety and psychological wellbeing. The pronounced vulnerability associated with early-morning sirens further suggests that targeted post-event mitigation strategies, such as delayed morning obligations or providing additional support following early-morning sirens, could help reduce their psychological impact.
On the clinical level, interventions could be offered during times of crises, as even low-intensity programs implemented during wartime have been shown to improve sleep, reduce stress symptoms, and enhance wellbeing (Kurapov et al., 2025; Yankovitch et al., 2025). Clinicians working with conflict-affected individuals should prioritize assessing sleep difficulties and consider scalable sleep-focused interventions, including psychoeducation, sleep hygiene, and low-threshold behavioral strategies designed to reduce nocturnal hyperarousal and facilitate return to sleep. Overall, these results emphasize the central role of sleep in times of crisis and offer guidance for protecting both physical safety and emotional health in populations exposed to war and other acute stressors.
FundingThis study was supported by the Israel Science Foundation (Grant 567/24) and The Charles, Evelyne and Sandra Dolansky Research and Treatment Program for Suicide Prevention. The funding organizations had no role in study design, data collection and analysis, manuscript preparation, or the decision to submit for publication.
Declaration of generative AI and AI-assisted technologies in the manuscript preparation processDuring the preparation of this work, the authors used ChatGPT-5.2 and Elicit in order to improve the writing style and guide with statistical analyses. After using these tools, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article.
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:
Michal Kahn reports financial support was provided by Israel Science Foundation. Michal Kahn reports financial support was provided by The Charles, Evelyne and Sandra Dolansky Research and Treatment Program for Suicide Prevention. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
We are deeply grateful to the participants who shared their experiences during this difficult period. We hope for peace, healthy sleep and wellbeing for all people throughout the region.
