Artificial Light at Night Consistently Impacts Avian Physiology and Behaviour: A Meta‐Analysis

ABSTRACT

Artificial light at night (ALAN) is a major anthropogenic pressure across the tree of life. While ALAN drives population declines in insects and reptiles, birds defy simple patterns. Avian responses to ALAN vary in direction and magnitude, yet the fitness consequences of these responses remain unresolved. We conducted a meta‐analysis to test whether ALAN alters avian physiology, behaviour and life‐histories, analysing 675 effect sizes from 36 studies across 30 species. We found consistent physiological and behavioural shifts under ALAN, while life‐history traits were unaffected. ALAN disrupted physiology through reduced sleep, higher metabolic rate and accelerated reproductive maturation. Behaviourally, daily activity was extended, with earlier onset, later offset, and increased nocturnal activity and foraging effort. Subset analyses revealed stronger ALAN effects in migratory species compared to residents and more pronounced in adults and females than in nestlings and males. Activity shifts, accelerated ageing and sleep disruption were amplified with brighter light. Birds appeared to buffer ALAN effects through physiological and behavioural adjustments, minimising impacts on life‐histories, a paradox likely explained by phenotypic plasticity or evolutionary adaptation. We identified key ecological vulnerabilities, emphasising the need for targeted conservation strategies in our increasingly illuminated world.

Artificial light at night (ALAN) is a major anthropogenic pressure across the tree of life, and avian responses are highly variable in both direction and magnitude. We used a meta‐analytical approach to disentangle ALAN effects across avian functional traits underpinning physiology, behaviour and life‐histories, while accounting for inter‐ and intraspecific variation and environmental context. Overall, we found consistent shifts in physiology and behaviour under ALAN, whereas life‐history traits remained unaffected.

1. Introduction

Artificial light at night (hereafter ALAN) has been increasing by ~2.2% annually, drastically transforming the predictable natural light–dark rhythms on Earth and becoming a major source of anthropogenic disturbance (Falchi et al. 2016; Gaston et al. 2014). Negative effects of ALAN have been documented across taxa, from microorganisms (Cesarz et al. 2023; Pu et al. 2019) and plants (Bennie et al. 2016; Matzke 1936) to invertebrates (Desouhant et al. 2019; Duarte et al. 2019; Underwood 2024), fish (Pulgar et al. 2023; Schligler et al. 2021), amphibians (Touzot et al. 2023; Wise 2007), reptiles (Kolbe et al. 2021; Taylor et al. 2022), birds (Adams et al. 2019; Xue et al. 2020) and mammals (Ditmer et al. 2021; Shier et al. 2020). In animals, physiological and behavioral responses to ALAN have been widely reported (Bumgarner and Nelson 2021; Cassone and Kumar 2022). However, whether these responses translate into consistent fitness and biodiversity consequences remains uncertain (Jägerbrand and Spoelstra 2023). While attraction to ALAN has been associated with increased mortality (Boyes et al. 2021; Bumgarner and Nelson 2021; Lao et al. 2020) and population declines (van Grunsven et al. 2020; Owens et al. 2020), evidence remains taxonomically limited. Mechanistically, ALAN can affect population persistence through physiological and behavioral disruption in insects (Boyes et al. 2021; Owens et al. 2020), disorientation of natural navigation in sea turtles (Dimitriadis et al. 2018; Leader et al. 2024) and fatal collisions with buildings in birds (Lao et al. 2020; Loss et al. 2019; Van Doren et al. 2021). Birds, however, exemplify more complex responses than simple attraction‐related risks and represent an important model for understanding ALAN effects across biological levels. Despite demonstrated vulnerability to ALAN in sleep, cognition, behavior, and other fitness‐related mechanisms (Aulsebrook et al. 2021; Dominoni 2015), population‐level responses varied markedly, with some species declining (Rodriguez et al. 2017), whereas others persisted or even thrived in illuminated environments (Wilson et al. 2021). This variation is particularly striking given that, unlike most vertebrates, many bird species are often abundant in urban habitats where nocturnal illumination is pervasive (Aulsebrook et al. 2021; Dominoni 2015).

A growing body of evidence suggests that ALAN significantly alters the interplay among physiological, behavioural, and life‐history traits determining a bird's capacity to maintain homeostasis with its environment (Capilla‐Lasheras et al. 2017; Sibly et al. 2012; Wingfield et al. 2017). Within these domains, specific functional traits (defined here as measurable characteristics of biological relevance) may respond directly to ALAN (Nock et al. 2016). A primary mechanism underlying responses to ALAN is disruption of the circadian clock, which synchronises biological functions with natural light–dark cycles (Cassone 2014) and is closely connected to immune regulation, stress physiology, endocrine function and metabolic pathways (Gotlieb et al. 2018; Jerigova et al. 2022; Markowska et al. 2017). ALAN can also alter the secretion of melatonin, a hormone produced during darkness to modulate sleep, metabolism and activity in diurnal birds (Bao et al. 2023; Wang et al. 2012). Consistent with these mechanisms, physiological responses to ALAN have been associated with reduced immunity, reflected by increased blood parasite loads in dark‐eyed juncos (Becker et al. 2020), and earlier endocrine reproductive activation through accelerated testes growth in adult great tits (Dominoni et al. 2018; de Jong et al. 2016). Strong evidence further indicates that ALAN alters behaviour and life‐history traits, such as extended daily activity and advanced reproductive timing across species (Pease and Gilbert 2025; Senzaki et al. 2020). Nonetheless, avian responses to ALAN varied across life stages, sexes, and ecological strategies. For instance, nestlings experienced higher energetic demands and greater physiological sensitivity than adults (Assadi and Fraser 2021; Oswald et al. 2018; Vézina and Salvante 2010). At adulthood, reproductive energy allocation and sensitivity to environmental stressors differed between sexes: males primarily invested in foraging and territorial activities, while females bore higher endocrinological costs linked to egg production and hormonal regulation (Sibly et al. 2012). ALAN exposure risks may also differ among species, with migratory birds and those occupying open habitats experiencing greater exposure (Adams et al. 2019; Bhagarathi et al. 2024; Burt et al. 2023). Remarkably, ALAN effects on specific avian traits are often context‐dependent or equivocal. For instance, sleep disruption has been documented only in breeding female great tits, or at light intensities exceeding 1.6 lx (Raap, Pinxten, and Eens 2016; Ulgezen et al. 2019). Many studies reported no impact of ALAN on reproductive output, such as clutch or brood size (Grunst et al. 2019; de Jong et al. 2015; Kempenaers et al. 2010; Ouyang et al. 2015), whereas others documented reduced hatching success (Assadi and Fraser 2021; Injaian et al. 2021; Malek and Haim 2019; Senzaki et al. 2020). Similarly, while some studies detected no changes in body mass under ALAN (Injaian et al. 2021; Kumar et al. 2021; Moaraf et al. 2021), others reported reduced nestling mass attributed to decreased parental care (Cianchetti‐Benedetti et al. 2018; Reid et al. 2025) and increased energy expenditure (Ferraro et al. 2020; Raap, Casasole, et al. 2016).

Overall, the impacts of ALAN on birds have been reviewed at the physiological and behavioural levels (Dominoni 2015; Grunst and Grunst 2023; Helm and Liedvogel 2024) or in terms of ecological outcomes, such as distribution, abundance and migratory routes (Adams et al. 2021; Jägerbrand and Spoelstra 2023). Sanders et al. (2020) provided meta‐analytical evidence of ALAN impacts across taxa, including birds, and classified responses into broad categories (e.g., physiology and activity patterns) and subcategories (e.g., gene expression and hormone levels). However, these categories encompass diverse biological functions through which ALAN may act, making it difficult to identify general mechanisms underlying avian responses or to determine how particular functional traits contribute to observed fitness and ecological outcomes. A synthesis that explicitly disentangles ALAN effects across functional traits underpinning physiology, behaviour, and life‐histories, while accounting for inter and intraspecific variation and environmental context, would therefore provide a more mechanistic understanding of its impacts on birds. Yet, such a study remains lacking. To address this gap, we conducted a systematic review and meta‐analysis to quantitatively synthesise the effects of ALAN on avian functional traits, providing a framework relevant to mitigation, evolutionary and conservation research. Based on available studies, we selected specific moderators to investigate whether ALAN affects physiology, behaviour, and life‐history traits. These three broad categories were divided into 16 functional traits covering various measurements of intraspecific (e.g., life stages and sexes) and interspecific traits [e.g., migratory status, habitat densities, study environment (captive vs. wild) and population trends]. Additionally, we tested whether ALAN was associated with disruption of circadian clock functioning by shifting the expression of core clock genes, particularly at night, with endocrine‐molecular downstream effects on ageing, immunity, metabolic rate, neuronal cognition, sleep regulation and reproductive maturation. These physiological imbalances could affect behavioural aspects, including activity offset and onset, foraging effort, levels of daily and nocturnal activity, and ultimately life‐history traits (body mass and size, reproductive phenology and success). Finally, we assessed ALAN's effects under varying experimental regimes (light intensities, days since ALAN exposure, day vs. night and tissue types) to identify the specific conditions under which ALAN exerts the strongest impact on avian performance and fitness.

2. Methods

2.1. Literature Review

We performed a systematic research of the published literature in three different databases: Scopus, Web of Science (All databases) and PubMed on the 28th of February 2023 via an institutional subscription at the Jagiellonian University, Kraków, Poland. We retrieved relevant literature on avian responses to ALAN using a specific search string (reported in Supporting Information Table 1). The search yielded 1507 records, including 1480 publications identified through our keyword search and 27 additional publications from citation search that met our inclusion criteria. All records were imported into Rayyan (Ouzzani et al. 2016). After removing duplicates (n = 379), 1128 unique studies remained and were selected for screening. The results of each search phase are reported in the Preferred Reporting Items for Systematic Reviews and Meta‐Analysis (PRISMA) diagram (Supporting Information, Figure S1).

TABLE 1.

The 16 avian functional traits underpinning avian performance and fitness, sorted by meta‐analytical categories: Physiology (red), behaviour (blue) and life‐history (violet). Functional traits are presented with a ‘Functional note’ describing their relevance in the context of avian performance. ‘Example proxies’ include the most representative variables from the published literature assigned to each trait. Variables expected to negatively impact the corresponding functional trait are marked with (‐). Circadian clock functioning (52 effect sizes) was excluded from the main model because data was supported by only two studies. The full list of effect size assessments is shown in Table S1.

Level	Avian functional trait	Functional note	Example proxy
Physiology	Ageing	Decline in physiological integrity over time affecting reproduction and survival	Relative telomere length, TAC‐total antioxidant activity (‐)
	Circadian clock functioning	Clock core genes regulating physiology and behaviour within a 24 h light–dark cycle	Nocturnal Bmal1, Per2 mRNA expression (‐)
	Immunity	Pathogen defence, vital for health and fitness	HP‐haptoglobin, % bacterial E. coli killing
	Metabolic rate	Biochemical processes for energy production and allocation	Daily energy expenditure (DEE), body temperature
	Neuronal cognition	Biochemical mechanisms underlying cognitive functions	Neuronal densities linked to neuronal plasticity
	Reproductive maturation	Endogenous timing, activation and development of reproduction	Luteinizing hormone (gonadal maturation)
	Sleep regulation	Biochemical modulators of sleep onset and duration, vital for behavioural functioning	Melatonin‐induces sleep onset, achm3‐awakens promoter gene (‐)
Behaviour	Activity offset	Time of day when activity ends	Last activity of the day and parental visit
	Activity onset	Time of day when activity begins	First nest entry or chorus time
	Foraging effort	Food acquisition strategies	Food intake and prey per minute
	Level of activity	Duration of diurnal behaviours linked to energetic expenditure	Activity duration and activity counts per hour
	Nocturnal activity	Extent of behavioural activity during night‐time hours	Sleep duration‐less activity (‐), awakening time
Life‐history traits	Body mass	Indicator of energetic reserves supporting survival and reproductive investment	Body mass, weight and growth rate
	Body size	Body dimensions shaping metabolism and reproductive capacity	Tarsus, wing, and rectrices lengths
	Reproductive phenology	Seasonal reproductive timing to match with favourable conditions	Incubation, nesting, and fledging dates
	Reproductive success	Outcome of breeding effort, directly related to fitness	Parental feeding rate, nest predation (‐)

2.2. Inclusion Criteria

The first screening involved reading the title and abstract to include studies using the following criteria: (1) studies reporting effects of ALAN on avian species in the field or laboratory; (2) including quantitative measurements of light (measured in lx); (3) reporting ALAN‐DARK paired comparisons, where the DARK condition corresponds to a control group exposed to natural nocturnal light levels (≤ 0.2 lx), and the ALAN group is exposed to ALAN up to 100 lx. Evidence indicates that animals respond to very low nocturnal light levels (≤ 0.1 lx), for example, great tits advance activity onset at 0.15 lx (Aulsebrook et al. 2022; Spoelstra et al. 2018). Nevertheless, we considered 0.2 lx as the maximum natural night illumination, as lunar light intensity can reach up to 0.26 lx during a super moon occurring once per year (Kyba et al. 2017). As for ALAN intensity levels, we aimed for studies reporting realistic real‐life values; streetlamps produce ~1.6 to 10 lx at ~10 m distance, with urban light pollution reaching maximal intensity of 100 lx on an illuminated stadium (Gaston et al. 2013; Rich and Longcore 2013; Sanders et al. 2020). We excluded literature reviews, conference reports and patents. We did not search for grey literature to standardise the data for a robust quantitative synthesis of peer‐reviewed studies and to allow contact with authors for additional data or methodological details. We also removed studies concerning meat/egg production or any pedigree‐bred birds to avoid including organisms artificially selected for specific traits, which could potentially influence their response to ALAN. Studies related to photoperiod, natural daylight, and those using urban areas as indirect proxies for artificial light were omitted. Studies comparing traits between urban (light) and rural populations were also removed to avoid potential confounding effects such as noise or human activity. Finally, we excluded data from within individual design studies, where the same group of birds was exposed to both ALAN and DARK conditions at different time points. We took this decision to homogenise the experimental designs of the studies included in this meta‐analysis and to avoid the potential influence of prior exposure to ALAN in the response to natural DARK conditions (and vice versa). The first screening was repeated in 10% of the studies by P. C‐L. and J.S., resulting in 95% of inclusion consistency. If information critical for inclusion or exclusion (e.g., type of light, light measurement, or experimental design) was not reported in the abstract, the article was assigned for full‐text screening. A description of the number of studies excluded at each screening stage and the exclusion reasons are shown in Table S7.

2.3. Full Screening and Data Extraction

The first screening yielded 180 studies that we read in full to extract the quantitative data needed to calculate effect sizes for ALAN–DARK paired comparisons. At this stage, studies were still assessed for compliance with our inclusion criteria and sufficient quantitative data availability (means, variance, and sample size), leading to the exclusion of 144 articles. Our final dataset comprised 36 studies, from which we collected the following moderators: publication year, life stage (adults or nestlings), sex, study environment (captivity or wild), focal species and light intensity. Since we focused on estimating the main effect of ALAN on avian functional traits, we excluded data from models reporting interactions (e.g., sex or age), unless the authors provided data to extract individual measures from them. We aimed to extract mean, standard deviation (SD), and sample size (n) from both ALAN and DARK groups. We extracted 675 effect sizes: 43 from tables and text, 242 from raw data provided in the published electronic Supporting Information and 390 from figures. We used ‘Web Plot Digitizer’ software (Tantry et al. 2021) to extract means and SD from the figures. We calculated 13 standard errors (SE) from two different studies, which were reported as 95% confidence intervals (Cooper et al. 2009). We used the formula: SD = SE× √n to calculate SD from 285 SE. In three studies, we calculated the mean and SD from eight quartiles and medians following Hozo et al. (2005). From four studies, we extracted the mean and SD from 17 estimates from linear mixed models reporting differences between ALAN and DARK groups (the method and formula are reported in the SI). We confirmed that effect sizes derived from raw data did not differ significantly from those obtained in models by testing the type of effect size (raw vs. model) as a moderator. As the type of effect size did not account for a significant proportion of variance, the variable was excluded from the final models. When the data reported in a study were incomplete (e.g., the mean was given, but the SD, SE, or sample size for one of the groups was missing), we contacted the authors. Of the 14 authors contacted, nine did not reply, while four provided additional information or raw data and are mentioned in the acknowledgements.

2.4. Mapping Effect Sizes Onto Avian Functional Traits

To provide a comprehensive framework for understanding avian responses to ALAN, we assigned the effect sizes to 16 functional traits underpinning physiological, behavioural and life‐history mechanisms of avian performance and fitness. Physiological traits included ageing, elements linked with accelerating (telomere shortening) or slowing (antioxidants) physiological decline over time; circadian clock functioning, as the expression of core clock genes; immunity, defence responses against pathogens (e.g., anti‐bacterial response); metabolic rate, factors contributing to energy production and expenditure; neuronal cognition, mechanisms underlying cognitive functions (e.g., neuronal density); reproductive maturation, that is, processes contributing to reproductive development and activation (e.g., gonadal maturation); and sleep regulation, referring to biochemical modulators of sleep onset and duration (e.g., melatonin levels). Behavioural functional traits comprised activity offset, as the time of end of daily activity; activity onset, as the time of start of daily activity, foraging effort, the investment in food‐searching and handling (e.g., preys per minute or food intake); level of activity, as the duration or amount of activity during the day (e.g., activity counts); and nocturnal activity (e.g., activity counts at night). Finally, life‐history traits encompassed body mass and size, as any phenotypical measure related to individual condition and fitness maintenance; reproductive phenology, as any seasonal timing of reproductive events (e.g., fledging date); and reproductive success, the outcome of breeding efforts (e.g., number of fledglings). We provide a detailed description of each functional trait in Table 1.

To minimise the risk of inconsistent labelling of effect sizes into functional traits, we based the assignment on the definitions provided in the original studies. Since some genes and physiological features may capture diverse biological functions, each effect size was assigned to a specific functional trait based on the main function targeted in the original study. To enable biological interpretation of the meta‐analysis results across studies, we standardised the direction of effect sizes within each functional trait. When inverse metrics were used to quantify the same biological response (e.g., oxidative damage vs. antioxidant levels for ageing), effect sizes were sign‐adjusted to ensure that positive values consistently represented an increase in the focal functional trait (e.g., accelerated ageing). Effect sizes were only multiplied by −1 when required to align metric directionality (e.g., antioxidant levels); no effect sizes were altered when the original sign already matched the functional interpretation (e.g., oxidative levels). Effect sizes that were multiplied by −1 and the rationale of their inversion are explained in Table S8. Although second clutches may exert varying pressures on parents in the context of ALAN, effect sizes for second clutches were available for only one study. We therefore included effect sizes for reproductive output only from first broods. Effect sizes reported across a 24‐h period were categorised as either ‘day’ (e.g., morning, noon) or ‘night’ (e.g., midnight) traits. We calculated the difference in mean and SD from eight studies (n = 67 effect sizes) reporting effect sizes before (baseline) and after ALAN exposure, where baseline corresponds to measurements obtained under natural light–dark or pre‐exposure conditions (details of the conversions are provided in the SI). We included functional traits in the global meta‐analytical model only if they were reported in at least three different studies. Circadian clock functioning was reported in only two studies and therefore excluded from the main model. A description of the assignment of effect sizes to functional traits and their directionality is provided in Table S1.

2.5. Data Analysis

All the analyses were conducted in R version 4.4.2.

2.6. Phylogeny and Interspecific Variation

To account for the phylogenetic variation in our meta‐analytic models, we computed a correlation matrix based on species relationships. This matrix was derived using branch lengths from Open Tree of Life phylogenies via the R package ‘rotl’ version 3.1.0 (Michonneau et al. 2016). The phylogenetic tree used for the global analysis is shown in Figure 1B. We also retrieved the migration status (migratory, partially migratory, or sedentary) and habitat density (dense, open or semi‐open) from the AVONET database (Tobias et al. 2022), and the population trend (decreasing, increasing, or stable) from the IUCN Red List of Threatened Species (https://www.iucnredlist.org) for each species. We assigned the activity pattern (diurnal or nocturnal) to each species and found that all species were diurnal, which reflects a focus on diurnal species in the available literature on ALAN effects in birds. A summary of interspecific traits is shown in Table S2.

FIGURE 1.

Geographic and phylogenetic distribution of the meta‐analytical dataset including 675 effect sizes reporting effects of artificial light at night on avian performance and fitness. (A) Global map distribution of the 36 studies included (number of studies per country in yellow), along with the species most represented in the dataset. All images are copyright‐free (CC BY‐SA) extracted from www.commons.wikipedia.org. Parus major (top left), Taeniopygia guttata (bottom left), Cyanistes caeruleus (top right) and Corvus splendens (bottom right). Authors: Alexis Lours, Mykola Swarnyk, Ken Billington and Dr. Raju Kasambe, respectively. (B) Phylogenetic tree of the 30 avian species included. For each species, k represents the number of effect sizes and bars show the proportion of effect sizes in physiology (red), behaviour (blue) and life‐history traits (purple). The table shows the 16 functional traits encompassing these three main categories. Circadian clock functioning (52 effect sizes) was excluded from the global models because data was supported by only two studies.

2.7. Effect Sizes

We calculated standardised mean differences between paired ALAN and DARK effect sizes using the function ‘escalc’ in the R package ‘Metafor’ version 4.6.0 to estimate Hedges's g and its sampling variance (Viechtbauer 2010). Our final dataset included 675 ALAN‐DARK paired comparisons from 30 species across 36 different studies published between 2006 and 2022. Circadian clock functioning (52 effect sizes) was excluded from the main model due to insufficient number of studies (n = 2) leaving 623 effect sizes in the global analysis.

2.8. Meta‐Analytical Models

Although the response variables encompassed diverse biological and statistical variation, we first fitted a phylogenetic multi‐level (intercept‐only) meta‐analysis with Hedges's g as the response variable including within and between‐study and phylogenetic variance components. This approach allowed us to estimate the overall heterogeneity in our data and the overall mean effect size across traits prior to assessing our moderators' explaining variation among effect sizes (Table 2; Model 1). Then, we assessed the directional influence of ALAN on each functional trait by including the functional trait category as a moderator (Table 2; Model 2). Second, we evaluated the influence of intraspecific variation by comparing the effects of ALAN on functional traits between life stages (adults vs. nestlings) and sexes (males vs. females). To enable direct comparisons, we ran separate models for adults (Table 2; Model 4) and nestlings (Table 2; Model 5) from subsets containing the same avian functional traits, and we followed the same approach for females (Table 2; Model 6) and males (Table 2; Model 7). Third, we tested the influence of interspecific variation by fitting a multi‐level model with migration and habitat density as moderators, and we repeated this model in a subset including only studies performed in the wild (Table 2; Model 8 a‐b). Since we observed moderate collinearity among moderators (GVIFs < 5) in this model, we used Type II tests, which evaluate each moderator's significance while accounting for the others. Then, because species traits were not available for all the species, we tested the influence of study environment and population trend in separate models (Table 2; Models 9 and 10). All the models testing the effects of intra and interspecific traits included avian functional traits as moderators.

TABLE 2.

The meta‐analytical models evaluating the effects of artificial light at night (ALAN) on avian performance and fitness. ‘Model index’ contains a series of sequential numbers from 1 to 10 (model IDs) to facilitate understanding of methods and the corresponding results. ‘Data’ refers to the effect size inclusion and ‘k’ represents the number of effect sizes for each model. ‘Moderators’ show the categories included as moderators in each model, ‘Details’ provides a short description of the model and ‘Output section’ shows where to find each model output.

Model index	Evaluating	Data	k=	Moderators	Details	Output section
1	Effects of ALAN on functional traits	All effect sizes	623	Intercept only	Overall meta‐analysis on the effects of ALAN on avian functional traits	Table 3; Figure 2
2			623	Functional trait‐1		Table 3; Figure 2
3	Publication bias		623	√ inverse of the sample size	a. Small study effects	Figure S7A
				Publication year (mean‐centred)	b. Time lag bias	Figure S7B
4	Influence of intraspecific variation	Effect sizes reported in adults	134	Functional trait‐1	Comparing ALAN effects between life stages	Table 4; Figure 3A
5		Effect sizes reported in nestlings	52	Functional trait‐1		Table 4; Figure 3A
6		Effect sizes reported in females	48	Functional trait‐1	Comparing ALAN effects between sexes	Table 4; Figure 3B
7		Effect sizes reported in males	104	Functional trait‐1		Table 4; Figure 3B
8	Influence of interspecific traits	Effect sizes from species reported in AVONET data	592	Migration‐1 + habitat density‐1 + functional trait‐1	Testing the influence of species traits on the effects of ALAN on functional traits	Table S3; Figure 4A,B
		Effect sizes only for studies performed in the wild from species reported in AVONET data	208	bMigration‐1 + habitat density‐1 + functional trait‐1		Table S9; Figure S9A,B
9		Effect sizes from captive and wild studies	623	Study environment + functional trait‐1		Table S3; Figure 4C
10		Effect sizes from species with population trend status	614	Population trend + functional trait‐1		Table S3; Figure 4D

To treat effect sizes from the same publication as non‐independent, we fitted phylogenetic multilevel meta‐analytical models by including study ID and individual observation ID as random effects. We also included species ID and the phylogeny as a random effect when the model included two or more different species. We used the function ‘i2_ml’ in the R package ‘OrchaRd’ version 2.1.3 to calculate the percentage of total relative heterogeneity and the heterogeneity explained by each of the random factors (Nakagawa et al. 2023).

2.9. Subset Analysis

Given that the circadian clock underlies physiological, behavioural, and life‐history changes induced by ALAN, we tested ALAN's effects on circadian clock function‐associated traits with an intercept‐only model and then with models including light intensity and time of day as moderators. Since light intensity was reported for most effect sizes, we evaluated its influence across overall functional traits. Then, we tested the influence of varying experimental regimes included in our dataset: light intensities (e.g., 0.5, 1, 5 lx), days since ALAN exposure, time of day (day vs. night), and tissue types (e.g., liver, spleen, brain) by grouping repeated effect sizes per condition. Individual subset analyses were then conducted for each functional trait. A summary of the models included in the subset analysis is shown in Table S4; Models 11–15.

2.10. Publication Bias

We examined the data for indicators of two types of publication bias, small study effects and time lag bias, following Nakagawa et al. (2023). We ran two additional multilevel meta‐analytical models, including as a single moderator either the square‐root of the inverse of the effective sample size or the mean‐centred year of study publication (Table 2; Model 3). We used the function ‘r2_ml’ in R package ‘OrchaRd’ version 2.1.3 to calculate the variation explained by these moderators (R ² _marginal) for each model (Nakagawa et al. 2022).

3. Results

3.1. Effects of ALAN on Avian Functional Traits

The intercept‐only model revealed a nonsignificant near‐zero effect of ALAN on the overall effect sizes (mean estimate [95% CI] = 0.056 [−0.484, 0.595]; Figure 2). Total heterogeneity was high (I ² = 89.9%), with most variance arising within studies (57.6%), among studies (16.5%) and phylogenetic effects (14.2%), while among‐species heterogeneity was low (1.64%). Model 2 (Table 3), including functional traits as a moderator, showed that daily avian activity started earlier (mean estimate [95% CI] = −1.496 [−1.910, −1.082]) and ended later (mean estimate [95% CI] = 0.666 [0.239, 1.093]) with light pollution. ALAN was associated with higher metabolic rate (mean estimate [95% CI] = 0.474 [0.029, 0.919]), accelerated reproductive maturation (mean estimate [95% CI] = 0.649 [0.029, 1.269]) and higher nocturnal activity (mean estimate [95% CI] = 0.650 [0.101, 1.200]), and foraging effort (mean estimate [95% CI] = 1.288 [0.646, 1.929]). ALAN was also associated with disrupted sleep regulation (mean estimate [95% CI] = −0.493 [−0.971, −0.016]). A non‐significant trend indicated that ALAN may increase the level of diurnal activity (mean estimate [95% CI] = 0.440 [−0.067, 0.948]), and no significant effects were detected for the remaining traits (full model outputs are shown in Table 3; Figure 2).

FIGURE 2.

Artificial light at night (ALAN) induced changes in specific physiological and behavioural functional traits. Results from the meta‐analytic model including a moderator of different functional traits (y‐axis) underpinning physiology (red dots), behaviour (blue dots) and life‐history (purple dots) traits. The grey figure below shows the mean effect size of ALAN across all functional traits fitted by a phylogenetic multi‐level (intercept only) meta‐analytic model. Both orchard plots show estimates for Hedges's g (x‐axis) along with its 95% confidence intervals (CIs, thick whisker) and 95% prediction intervals (thin whisker). Positive and negative estimates indicate positive and negative effects of ALAN, respectively. The asterisks show the significance level (*p < 0.05, **p < 0.01 and ***p < 0.001) and k represents the number of effect sizes for each level, while the number of independent studies is shown in brackets. Circadian clock functioning (k = 52) was excluded from the main model because data was supported by only two studies. Outputs of full models 1 and 2 are shown in Table 3.

TABLE 3.

The global meta‐analytical models on the effects of artificial light at night (ALAN) on avian functional traits underpinning physiology (red), behaviour (blue) and life‐history traits (violet). Intercept only model displays the entire model summary with the associated heterogeneity (I ²) explained for each of the random factors, I ² total (total heterogeneity), I ² ACC (study observation), I ² sID (individual effect size observation), I ² species ID (species) and I ² phylogeny (phylogenetic correlation between species). Bold estimates indicate confidence intervals (CIs) that did not overlap zero. For each level, k (number of effect sizes), n (number of studies) and species (number of species) are reported. The global model included avian functional traits as a moderator (15 levels, k = 623). Circadian clock functioning was excluded due to insufficient number of studies.

Model 1		Mean estimate [95% CI]	k	n	Species
Intercept only model		0.056 [−0.484, 0.595]	623	36	30
I 2 total = 89.91
I 2 ACC = 16.51
I 2 sID = 57.55
I 2 species ID = 1.64
I 2 phylogeny = 14.21
Model 2 Avian functional traits
Physiology	Ageing	0.383 [−0.095, 0.861]	36	6	3
	Immunity	0.271 [−0.228, 0.769]	36	5	3
	Metabolic rate	0.474 [0.029, 0.919]	58	9	5
	Neuronal cognition	0.095 [−0.415, 0.605]	60	5	3
	Reproductive maturation	0.649 [0.029, 1.269]	69	3	3
	Sleep regulation	−0.493 [−0.971, −0.016]	39	9	4
Behaviour	Activity offset	0.666 [0.239, 1.093]	78	8	8
	Activity onset	−1.496 [−1.910, −1.082]	111	12	14
	Foraging effort	1.288 [0.646, 1.929]	12	3	8
	Level of activity	0.440 [−0.067, 0.948]	26	6	3
	Nocturnal activity	0.650 [0.101, 1.200]	19	4	6
Life‐history traits	Body mass	0.359 [−0.152, 0.869]	19	13	6
	Body size	0.032 [−0.533, 0.597]	12	6	3
	Reproductive phenology	0.323 [−0.294, 0.940]	11	5	5
	Reproductive success	0.037 [−0.423, 0.496]	39	9	5

3.2. Influence of Intraspecific Variation

When analysing adults and nestlings separately, we found that ALAN‐induced changes in functional traits in adults but not in nestlings (Model 4–5; Table 4). Adults exposed to ALAN showed increased body mass (mean estimate [95% CI] = 1.256 [0.580, 1.932]), accelerated ageing (mean estimate [95% CI] = 0.568 [0.017, 1.12]) and disrupted sleep regulation (mean estimate [95% CI] = −0.476 [−0.879, −0.072]). Models assessing sex‐specific effects (Models 6–7; Table 4) showed that ALAN advanced activity onset in males (mean estimate [95% CI] = −2.393 [−3.831, −0.955]), but not in females (mean estimate [95% CI] = 0.096 [−0.547, 0.740]). Conversely, ALAN induced reproductive maturation (mean estimate [95% CI] = 0.881 [0.225, 1.536]), enhanced neuronal cognition (mean estimate [95% CI] = 1.111 [0.236, 1.986]) and increased foraging effort (mean estimate [95% CI] = 1.235 [0.397, 2.073]) in females, with no comparable changes in males. Full model outputs are shown in Table 4; Figure 3A,B.

TABLE 4.

Influence of intraspecific variation on the effects of light pollution on avian functional traits. We compared effects between life stages (adults vs. nestlings) by running separate subset models for adults and nestlings. To enable direct comparisons, subsets included only functional traits reported in both life stages. The same approach was applied to compare effects between sexes (females vs. males). The table shows the results of each meta‐analytic model with bold estimates indicating confidence intervals (CIs) that did not overlap zero. For each level, k represents the number of effect sizes and n the number of studies.

Functional trait	Estimate [95% CI]	k	n	Estimate [95% CI]	k	n
	Adults Model 4			Nestlings Model 5
Ageing	0.568 [0.017, 1.12]	17	3	−0.187 [−0.750, 0.375]	19	3
Immunity	0.167 [−0.416, 0.749]	20	2	0.023 [−0.539, 0.585]	16	3
Metabolic rate	0.378 [−0.006, 0.762]	53	7	0.133 [−0.506, 0.772]	3	2
Sleep regulation	−0.476 [−0.879, −0.072]	36	7	−0.488 [−1.156, 0.180]	3	2
Body mass	1.256 [0.580, 1.932]	8	5	−0.240 [−0.794, 0.313]	11	8
	Females Model 6			Males, Model 7
Ageing	−0.092 [−0.711, 0.526]	7	3	0.465 [−0.996, 1.927]	17	3
Immunity	0.207 [−0.435, 0.849]	8	2	−0.054 [−1.504, 1.395]	24	2
Metabolic rate	0.437 [−0.084, 0.958]	11	2	0.144 [−1.302, 1.590]	39	4
Neuronal cognition	1.111 [0.236, 1.986]	6	1	0.280 [−1.226, 1.785]	29	2
Reproductive maturation	0.881 [0.225, 1.536]	14	1	0.214 [−1.319, 1.747]	55	2
Sleep regulation	−0.458 [−1.001, 0.085]	11	3	−0.679 [−2.224, 0.865]	7	3
Activity offset	0.591 [−0.066, 1.248]	8	1	0.630 [−0.805, 2.065]	48	1
Activity onset	0.096 [−0.547, 0.740]	8	1	−2.393 [−3.831, −0.955]	53	2
Foraging effort	1.235 [0.397, 2.073]	5	1	2.519 [−0.363, 5.401]	1	1
Level of activity	−0.087 [−1.313, 1.139]	2	1	0.325 [−1.139, 1.789]	16	2
Body mass	0.305 [−0.475, 1.086]	5	4	0.712 [−0.836, 2.260]	5	4
Body size	0.174 [−0.982, 1.331]	2	2	0.446 [−1.177, 2.069]	2	2
Reproductive success	0.292 [−0.238, 0.821]	11	3	−0.128 [−1.638, 1.382]	8	3

FIGURE 3.

Artificial light at night (ALAN) induced changes in specific functional traits only in adults and females. Results from meta‐analytic models comparing the effects of ALAN between (A) life stages (adults vs. nestlings) and (B) sexes (females vs. males). Avian functional traits underpinning physiology (red dots), behaviour (blue dots) and life‐histories (purple dots) are shown in y‐axis. The grey figures below show the mean effect size of ALAN across all functional traits. The orchard plots show estimates for Hedges's g along with its 95% confidence intervals (CIs, thick whisker) and 95% prediction intervals (thin whisker) in x‐axis. Positive and negative estimates indicate positive and negative effects of ALAN, respectively. The asterisks show the significance level (*p < 0.05, **p < 0.01 and ***p < 0.001), k represents the number of effect sizes, and the number of independent studies is shown in brackets for each level. Outputs of full models 4–7 are shown in Table 4.

3.3. Influence of Inter‐Species Traits

We found that ALAN particularly affected fully migratory species (Model 8; mean estimate [95% CI] = 1.088 [0.357, 1.818]), while the effects of ALAN were not significant for partial migrants (mean estimate [95% CI] = 0.879 [−0.004, 1.761]), and sedentary species (mean estimate [95% CI] = 0.737 [−0.171, 1.645]) (Figure 4A). Habitat density had no significant influence on light pollution impact (Model 8a; Figure 4B), but there was a tendency for ALAN effects to be negative in species from open habitats (mean [95% CI] = −0.715 [−1.469, 0.039]), and no effects were detected in those from semi‐open or dense habitats. Yet, the subset analysis for studies in the wild showed that ALAN had significant negative effects in species from open habitats, whereas no effects were detected in dense habitats (Model 8b; Table S9; Figure S8). In Model 8a, we used Type II tests to evaluate each moderator's significance while accounting for moderate collinearity among them (GVIFs < 5). All moderators significantly contributed to heterogeneity in effect sizes: migration (QM = 8.78, df = 3, p = 0.032), habitat density (QM = 229.64, df = 3, p < 0.0001) and functional trait (QM = 54.44, df = 13, p < 0.0001). Model 9 for study environment (captive vs. wild; Figure 4C) and Model 10 for population trend (decreasing, stable, increasing; Figure 4D) did not explain variation in ALAN effects. A summary of all models for interspecific traits is provided in Table S3.

FIGURE 4.

The effects of artificial light at night (ALAN) were stronger on avian functional traits of migratory species, while other interspecific traits had no significant influence. Results from the meta‐analytic models including: (A) migration, (B) habitat density, (C) study environment and (D) population trend as moderators, along with avian functional traits. The orchard plots show estimates for Hedges's g along with its 95% confidence intervals (CIs, thick whisker) and 95% prediction intervals (thin whisker) in the x‐axis, and the categories of each interspecific trait in the y‐axis. Positive and negative estimates assume a positive or negative effect of ALAN, respectively. The asterisks show the significance level (**p < 0.01), and k represents the number of effect sizes for each level, while the number of independent studies is shown in brackets. Outputs of full models 8–10 are shown in Table S3.

3.4. Subset Analysis on the Circadian Clock and Experimental Regimes

Model 11 for circadian clock functioning showed significant overall negative effects of ALAN. These effects were significant only if measured at night, and light intensity moderated the magnitude of these effects (details provided in Table S5; Figure S2A–C). While we did not find significant effects of light intensity on overall functional traits (Model 12a; Table S6), specific‐trait models revealed that higher light intensity was linked to accelerated ageing, reduced sleep regulation, earlier activity onset and increased nocturnal activity (Figure S3B,G,I,K). Longer ALAN exposure correlated with enhanced reproductive success (Model 13; Figure S4C). Positive effects of ALAN on neuronal cognition occurred only during the day, whereas ALAN‐induced reproductive maturation was independent of sampling time (Model 14; Figure S5D,E). Notably, increased foraging effort and sleep disruption, previously observed in the global model, were significant only at night (Model 14; Figure S5F,G). ALAN accelerated ageing in the liver but not in the hypothalamus, and reduced immunity was observed in the hypothalamus but not in the spleen (Model 15; Figure S6A,B). Detailed model outputs on experimental regimes are shown in Table S6.

3.5. Publication Bias

The dataset showed evidence of small‐study effects, indicating that studies with smaller sample sizes tend to report larger and more significant effects than those with larger samples (mean estimate [95% CI] = 1.025 [0.187, 1.864]; Figure S7A), though this effect only explained a low proportion of the variance (R ² _marginal = 0.023). Following Nakagawa et al. (2022), the intercept from this small‐study effects model provided an unbiased estimate of the overall effect size (intercept mean estimate [95% CI] = −0.315 [−1.011, 0.380]). This unbiased estimate was lower than the intercept‐only model estimate (Model 1; mean estimate [95% CI] = 0.056 [−0.482, 0.593]), suggesting an inflation of the naïve estimate towards positive values. However, both estimates overlapped considerably and included zero, indicating that while small‐study effects were detected, they did not substantially alter the overall conclusions. Conversely, we did not find evidence of time lag bias (mean estimate [95% CI] = 0.054 [−0.006, 0.113]; Figure S7B), the model showed a limited proportion of variance explained by time lag bias (R ² _marginal = 0.029).

4. Discussion

This meta‐analysis revealed consistent shifts in avian physiological and behavioural functional traits under ALAN. In order of effect magnitude: ALAN most strongly advanced daily activity onset, followed by increased foraging effort, delayed daily activity offset, accelerated reproductive maturation, elevated nocturnal activity, disrupted sleep regulation and increased metabolic rate. Light pollution increased body mass, accelerated ageing and reduced sleep regulation in adults, but not in nestlings. While ALAN increased foraging effort, improved neuronal cognition and accelerated reproductive maturation in females, males largely advanced their daily activity onset. Interestingly, light pollution had stronger effects on the performance of migratory birds, compared to partial migrants and sedentary species. These effects were phylogenetically heterogeneous but consistent across study environments, habitat openness and species' extinction risk.

The intercept‐only model showed no overall effect of ALAN. This is not surprising given that various functional traits influence avian responses in opposing directions (e.g., increased nocturnal activity with ALAN reflects a behavioural intensification, whereas accelerated ageing negatively impacts lifespan and fitness). High heterogeneity in Hedges's g indicates that ALAN effects are strongly trait‐ and context‐dependent, with most variance arising from within‐study variation and phylogenetic relationships among species. This level of heterogeneity is characteristic of ecological meta‐analysis integrating responses across species, biological traits, and experimental frameworks (Nakagawa et al. 2015; Senior et al. 2016). Looking across all functional traits, ALAN advanced avian onset activity into the early morning and extended offset activity into the evening, indicating a broad temporal expansion of activity. This behavioural shift was followed by increased foraging effort, particularly at night (Figure S5G) and by elevated nocturnal activity, suggesting a partial behavioural adaptation to illuminated conditions. Earlier onset and extended daily activity are among the most frequently reported behavioural responses to light pollution (Da Silva et al. 2016; Da Silva and Kempenaers 2017). We found that the earlier start of activity under ALAN was specific to males (Table 4; Figure 3B). Advanced activity onset is thought to provide potential benefits for male singing and extra‐pair opportunities (Dominoni and Nelson 2018; Kempenaers et al. 2010; Murphy et al. 2008), and has been widely discussed in the context of sexual selection in human‐altered environments (Capilla‐Lasheras et al. 2025; Charmantier et al. 2024; Halfwerk and Slabbekoorn 2015). This suggests that ALAN, as a major anthropogenic pollutant characteristic of urban environments, may amplify traits already under sexual selection. However, reproductive success also depends on female availability and reproductive strategies (Brouwer and Griffith 2019; Westneat and Stewart 2003). Recent work has revealed that urbanisation and ALAN can also advance activity timing in females (Capilla‐Lasheras et al. 2025; Champenois et al. 2026; McGlade et al. 2023; Womack et al. 2023), indicating that the earlier literature was predominantly male‐focused and underscoring the need to include female responses in future research.

Beyond sex‐specific responses, ALAN disrupts circadian rhythms through changes in core clock gene expression and melatonin suppression, which can shift daytime activity and promote increased nocturnal activity (Brandstätter 2002; Cassone 2014; Dominoni et al. 2022). Additionally, as ALAN improves nocturnal vision, it may stimulate birds to forage longer into the night, potentially increasing interspecific competition and predation risk (Bonter et al. 2013; Leveau 2020). Sleep disruption under ALAN (Figure 2), particularly at night (Figure S5F), is another trait possibly linked to increased nocturnal activity. This is not surprising in diurnal species and can connect physiological and behavioural responses to ALAN (table 3; Figure 2) through nocturnal melatonin suppression (Bentley 2001; Cassone 1990; Grubisic et al. 2019). Melatonin, central to initiating and maintaining sleep, was consistently suppressed under ALAN in our study. This aligns with extensive evidence reported across vertebrates (Cassone and Kumar 2022; Grubisic et al. 2019; Grunst and Grunst 2023), including recent meta‐analytical research (Yang et al. 2024). We also observed a dysregulation of molecular modulators of sleep onset (e.g., sik3 mRNA expression), and wakefulness promotion (e.g., camkII mRNA expression and oxalate). These modulators are also interconnected with metabolic processes (Batra et al. 2020; He et al. 2023; Li et al. 2023), likely reinforcing the link between disrupted circadian rhythms, poor sleep, and altered energy balance. Importantly, this overall expansion of activity likely incurred physiological costs, as increased locomotor effort elevates energy demands and metabolic rate (Batra et al. 2019). This is supported by our finding of higher metabolic rate under ALAN (Table 3; Figure 2). Metabolic rate is interconnected not only with activity levels but also with elevated corticosterone release as a co‐effect of ALAN‐induced melatonin suppression (Jimeno and Verhulst 2023; Michael Romero 2002; Sapolsky et al. 2000). Finally, we observed accelerated reproductive maturation under ALAN, particularly in endocrine markers that trigger breeding. While this trait falls outside the immediate behavioural‐metabolic pathway, it is functionally tied to the circadian system and melatonin‐mediated photoperiodic regulation (Helm et al. 2024; Helm and Liedvogel 2024). We detected that ALAN induced reproductive maturation by releasing reproductive hormones and increasing expression of reproduction‐related genes (Tables 1 and 3; Figure 2). These effects were specific to females (Table 4; Figure 3B), with no significant effects detected in males. Female reproductive physiology is likely more sensitive to environmental cues (e.g., photoperiod, stress and temperature), as reproductive timing and hormonal cycles depend on circadian and seasonal regulation (Helm et al. 2017). Such female‐related effects could also arise because males' reproductive activation is generally less sensitive to physiological changes (Ball and Ketterson 2007; McEwen and Wingfield 2003). Thus, elevated reproductive readiness under ALAN may carry long‐term fitness consequences, including mismatches with optimal breeding conditions and increased energetic demands on females. Notably, only females exhibited enhanced neuronal cognition and foraging effort under ALAN (Table 4; Figure 3B). This sex‐specific sensitivity may reflect females' greater physiological and behavioural investment during reproduction, particularly in egg formation, incubation, and provisioning (Love et al. 2005; Vézina and Salvante 2010). Consequently, females are likely more responsive to photoperiodic cues like ALAN that alter timing and energy allocation.

We did not detect any overall effects of ALAN on immune traits. Such effects often emerge in combination with additional stressors, such as pathogen exposure or immune suppressants (Markowska et al. 2017; Tognini et al. 2018; Ziegler et al. 2021), which were not consistently reported across our dataset. ALAN effects on avian immunity are also tissue‐dependent (Dominoni et al. 2022), which aligns with the tissue‐dependent patterns observed in our subset analysis (Table S6; Figure S6B). Additionally, several studies assessed immune traits across different tissues and light intensities using small sample sizes, which may have introduced small‐study effects (Figure S7A) and contributed to the limited evidence for ALAN's impact on immunity.

We found life stage‐specific differences in avian responses to ALAN. Adults displayed accelerated ageing, particularly in terms of the expression of ageing‐related genes (Table S1). The expression of genes encoding key antioxidant enzymes (gst, sod3, cat1, sirt1) and promoting mitochondrial biogenesis (IGF1) is remarkably sensitive to ALAN due to their tight connection to the circadian clock (Batra et al. 2022; Dominoni et al. 2022; Taufique et al. 2018). However, despite these molecular traits are indicators of ageing, direct ecological evidence linking ALAN to reduced lifespan (e.g., via survival probabilities or lifetime reproductive span) remains lacking. ALAN increased body mass in adults (Table 4; Figure 3A), which may result from greater energy intake through disrupted endocrine hunger regulation (Mahdavi et al. 2025) and extended foraging behaviour (Da Silva et al. 2016; de Jong et al. 2016; Titulaer et al. 2012). Our dataset included only one study reporting both body mass and foraging effort, and no studies have simultaneously studied endocrine hunger responses (e.g., ghrelin hormone levels) under ALAN. This calls for future research integrating these traits within the same biological system to better resolve their interrelated responses to ALAN. The absence of sleep disruption in nestlings, compared to clear ALAN effects in adults, suggests that ALAN vulnerabilities emerge later in life, once circadian, endocrine and behavioural systems are fully developed. Yet, this also raises the question of how circadian endocrine rhythms emerge during early life, which remain poorly understood despite evidence of circadian activity before hatching (Wellard et al. 2025). ALAN may therefore act as an age‐dependent stressor, with stronger consequences for adult maintenance and senescence pathways. On the interspecific level, we confirmed that the overall effects of ALAN on functional traits are significantly higher in long‐distance migrants compared to partial migrants and sedentary species (Table S3; Figure 4A). Migratory species rely on photoperiodic cues to regulate migratory timing and physiological preparation, making them likely more sensitive to ALAN (Delmore et al. 2020; Helm and Liedvogel 2024; Rowan 1930). While residents of light‐polluted areas might experience chronic and cumulative ALAN exposure over the year, nocturnal migrants often breed in the same locations as residents and may encounter more ALAN throughout their annual cycle by crossing areas with varying light levels (La Sorte and Horton 2021). Intermittent ALAN exposure in migrants may, therefore, force repeated shifts in physiological and behavioural states, imposing other costs than chronic exposure in residents. Importantly, migratory birds are strongly attracted to urban lights during stopovers (Burt et al. 2023; Cabrera‐Cruz et al. 2018; McLaren et al. 2018), which can be particularly detrimental for nocturnal species, who account for the majority (~60%) of collisions with illuminated buildings (Lao et al. 2020; Uribe‐Morfín et al. 2021). Hence, the pronounced responses of migrants to ALAN make them a priority for conservation measures and ecological monitoring. Our analysis showed a marginally non‐significant negative effect of open habitats on the overall impact of ALAN (Figure 4B), consistent with previous findings of smaller clutches in birds exposed to ALAN from open versus close habitats (Senzaki et al. 2020). Although study environment (captive vs. wild) did not significantly influence ALAN effects (Figure 4C), our subset analysis only from wild studies, indicated a significant negative ALAN effect on species preferring open habitats (Table S9; Figure S8B). Thus, the overall lack of habitat density effect (Figure 4B) may be influenced by captive studies, which lack natural environmental buffering (e.g., vegetation cover, weather, or behavioural avoidance) present in the wild and underscore the need of further research to capture more comprehensive habitat‐specific ALAN responses in natural populations. Finally, our analysis showed that species' population trends (increasing, decreasing, or stable, IUCN) did not influence ALAN responses (Figure 4D). This is unsurprising, as IUCN categories operate at broad spatial scales across multiple dimensions of conservation concern, whereas ALAN impacts often occur at local population scales. Recent evidence showed that species abundance and locations influence populations declines (Johnston et al. 2025), suggesting that finer‐scale trends could better capture avian vulnerabilities to anthropogenic stressors like ALAN. Our dataset, while taxonomically broad, included only 30 avian species (mostly passerines) distributed across three continents (Figure 1A), limiting our ability to capture the full ecological and population‐level diversity. We therefore encourage future research incorporating fine‐scale avian population declines susceptibility and ecologically relevant descriptors (e.g., latitudinal range extent) to better link ALAN effects to regional population dynamics.

Although we did not find a significant overall effect of light intensity across all functional traits (Table S6), our subset analysis showed that higher light intensities were associated with enhanced nocturnal activity, advanced daily activity onset, accelerated ageing and reduced sleep regulation (Table S6; Figure S3B,G,I,K). These amplified effects likely result from increased disruption of melatonin production and circadian gene expression. Brighter light more powerfully suppresses the nocturnal signals that regulate sleep, metabolic rate and cellular repair, which are key processes involved in ageing and daily behavioural rhythms (Dominoni et al. 2022; de Jong et al. 2016; Moaraf, Heiblum, et al. 2020; Moaraf, Vistoropsky, et al. 2020). From Figure S3B,G,I,K, we observed that the magnitude and direction of these effects became more pronounced above ca. 2 lx, suggesting a potential shift in nocturnal activity, daily activity onset, ageing and sleep regulation at higher light intensities. Nevertheless, this pattern should be interpreted cautiously, as it is based on a smaller subset of effect sizes, species, and functional traits and may not reflect general ALAN effects on avian performance and fitness. Importantly, physiological responses to lower light intensities (e.g., 0.5 lx) are well documented in laboratory‐captive experiments, where light intensity is continuous and spectrally simplified (Raap, Casasole, et al. 2016). Yet, free‐living individuals may be more sensitive to the light intensity heterogeneity present in natural environments, where illumination fluctuates spatially and temporally across microhabitats due to vegetation cover, habitat structure, weather and behavioural avoidance (Bennie et al. 2016; Dominoni et al. 2016; Kyba et al. 2017). While ALAN was here defined as illumination exceeding natural nocturnal darkness (below 0.2 lx under full moonlight) (Gaston et al. 2015; Sanders et al. 2020), anthropogenic lighting typically forms strong spatial gradients, reaching ~30 lx beneath streetlights and decaying rapidly to ca. 2 lx at 20 m distance (Bennie et al. 2016). Because light intensity declines rapidly with distance following the inverse‐square law, there are practical opportunities to mitigate ALAN impacts on wildlife. Although higher street illumination on major roads and high‐crime areas supports crime reduction and public reassurance (Welsh et al. 2022), more modest lighting is possible on smaller streets, residential zones, and urban parks. Field studies showed that increases from 0.1–0.7 to 0.5–1.4 lx can enhance perceived safety (Ishii Okuda and Fukagawa 2008; Vrij and Winkel 1991), suggesting that dimming remains compatible with human security. Applying the inverse‐square law, dimming typical 50W LED lamps to 25W reduces the distance at which light exceeds 2 lx by ~30% (from ~14 to 10 m), and doubling lamp spacing reduces cumulative exposure at midpoints by ~75%, creating darker refugia for urban wildlife without compromising safety.

We also found that the effects of ALAN on increased foraging effort and sleep disruption were only evident when measured at night, with no difference during the day (Figure S5F,G). Improvements in neuronal cognition were observed only during the day (Figure S5D), which is expected given the negative impact of disrupted sleep on nocturnal brain function (Grubisic et al. 2019; Moaraf, Vistoropsky, et al. 2020). Reproductive maturation, a long‐term physiological state governed by seasonal hormonal changes (Dominoni et al. 2013; Helm et al. 2024), was consistently upregulated in measurements taken during both, day and night (Figure S5E). This indicates that, unlike traits with strong daily fluctuations (e.g., circadian genes and melatonin levels), reproductive maturation traits do not vary within a single day yet still reflect a persistent impact of ALAN on long‐term reproductive physiology. Additionally, we observed that reproductive success increased with time since ALAN exposure (Figure S4C), possibly reflecting recovery from circadian disruption (Helm et al. 2024). Nevertheless, this pattern is based on one study with eight effect sizes and warrants cautious interpretation.

Circadian clock functional traits were excluded from our global analysis due to insufficient literature. Yet, our subset analysis revealed consistent ALAN‐induced shifts in circadian clock gene expression (e.g., Per2, Cry 1 and 4, ck1E and Bmal1), particularly at night when natural dark cycles are disrupted (Table S5; Figure S2C). This disruption may be underlying other physiological and behavioural responses observed in this study (Table 3; Figure 2). Such mechanism has been supported by recent reviews (Grunst and Grunst 2023; Helm and Liedvogel 2024; Wellard et al. 2025) and experimental studies showing that ALAN‐induced Bmal1 phase advancement is associated with earlier activity onset in free‐living house sparrows ( Passer domesticus ) and captive zebra finches ( Taeniopygia guttata ) (Coker et al. 2025; Hui et al. 2025). Together, these results stress the need for expanded research on circadian clock gene dynamics and their connections with other physiological and behavioural responses, particularly in wild populations, to better understand the mechanisms underlying avian responses to ALAN.

Our results did not allow us to support the notion that light pollution constrains avian life‐history strategies via functional traits (Figure 2). Specifically, we found no consistent overall changes in body size, reproductive phenology, or reproductive success. This contrasts with meta‐analytical evidence linking other anthropogenic disturbances (e.g., urbanisation and climate warming) to clutch size declines, advanced breeding phenology, and reduced sexual signal quality (Capilla‐Lasheras et al. 2022; Halupka et al. 2023; Sepp et al. 2018). Similarly, Senzaki et al. (2020) reported that ALAN increased reproductive success in some North American birds, particularly those with enhanced low‐light vision, which showed earlier laying dates and reduced clutch failure. Notably, reproductive phenology (k = 12) and success (k = 39) were represented by effect sizes from only five species in our analysis. While these effect sizes span multiple studies, years and experimental settings across a light intensity gradient in the northern hemisphere (Figure 1A), they do not provide evidence for a general influence of ALAN on life‐history traits. Instead, ALAN's effects appear to be life stage related, with higher body mass observed in adults but no detectable effects in nestlings (Figure 3A). Birds seem to compensate for ALAN‐induced stress through physiological and behavioural adjustments, prioritising core reproductive and survival functions, potentially through phenotypic plasticity (Hau and Goymann 2015) or evolutionary adaptation (Martin 2004; Sepp et al. 2018). Determining the mechanistic basis and long‐term sustainability of these compensatory responses represents a critical frontier for understanding avian resilience to anthropogenic light pollution.

5. Conclusions

Our meta‐analysis showed that avian species responded to ALAN through behavioural (extended daily activity and increased foraging effort) and physiological (elevated metabolic rate, sleep disruption and induced reproductive maturation) pathways. This aligns with previous evidence reporting that anthropogenic perturbations disrupt internal homeostasis, shifting energy acquisition‐metabolic expenditure trade‐offs with downstream behavioural consequences. Conversely, life‐history traits related to growth, reproductive success and survival appeared unaffected. These results supported the paradox often observed in light‐pollution research: while ALAN clearly alters physiology and behaviour, individual fitness and population‐level effects are more complex and context‐dependent. Birds likely adapt to light‐polluted environments, whether through phenotypic plasticity or genetic change: an encouraging indication for future studies of resilience across many species. Our subset analyses suggested that ALAN affects life stages and sexes differently, but these findings were based on limited representation and require cautious interpretation. Importantly, free‐living birds can behaviourally avoid ALAN, yet they are likely sensitive to the spatial and spectral heterogeneity of high‐intensity ALAN in natural environments. Many avian responses to ALAN remain poorly studied across functional traits and species, offering valuable opportunities for future research using rigorous designs that will strengthen cross‐study comparability and meta‐analytic synthesis.

Author Contributions

Sayuri Diaz‐Palma and Joanna Sudyka conceived the study. Sayuri Diaz‐Palma performed the literature search, literature screening and extracted effect sizes with input from Pablo Capilla‐Lasheras, Davide Dominoni and Joanna Sudyka. Pablo Capilla‐Lasheras and Joanna Sudyka validated effect size extraction. Sayuri Diaz‐Palma performed all statistical analyses with substantial advice from Pablo Capilla‐Lasheras. Sayuri Diaz‐Palma wrote the first draft of the manuscript, Mariusz Cichoń and Joanna Sudyka reviewed the first version. All authors contributed substantially to further revisions of the manuscript.

Funding

The study was funded by the NCN grant, SONATA no. 2019/35/D/NZ8/00889 awarded to JS. PC‐L was supported by a Global Marie Skłodowska‐Curie Actions Fellowship (HORIZON‐MSCA‐2023‐PF‐01‐101150591). The open‐access publication of this article was funded by the programme ‘Excellence Initiative–Research University’ at the Faculty of Biology of the Jagiellonian University in Kraków, Poland.

Supporting information

Table S1: The effect sizes included in the meta‐analysis on the effects of artificial light at night on avian functional traits. The table shows the effect sizes assigned to 16 avian functional traits underpinning avian performance and fitness, sorted by meta‐analytical categories: physiology (red), behaviour (blue) and life history traits (violet). Effect sizes are presented with a Functional note explaining its inclusion to the specific Functional trait. To enable interpretation of the meta‐analysis results, we assigned a direction to each effect size. If an increase in the effect size was expected to negatively impact the particular avian functional trait, the direction was negative. When the direction was not reported in the original study, effect sizes expected to have a negative impact were multiplied by −1. Circadian clock functioning (52 effect sizes) was excluded from the main model because data was supported by only two studies. The table summarises 675 effect sizes from ALAN–DARK paired comparisons across 36 studies published between 2006 and 2022.

Table S2: Summary of the inter‐species traits included in the meta‐analysis. The table shows the Scientific names of the 36 species included in our dataset followed by inter‐species traits: Migration, Habitat density (extracted from AVONET database), Population trend (collected from the IUCN Red List) and Activity pattern (diurnal or nocturnal). NA indicates missing data for this particular species.

Table S3: Summary of the meta‐analytic mixed‐effects models showing no significant influence of inter‐species traits on the impact of artificial light at night on avian functional traits. The table show estimates and confidence intervals (CIs) for each category of the species‐specific trait investigated. The first model included: migration and habitat density. The second and third models included study habitat and population trend as moderators. For each level, k represents the number of effect sizes, n the number of studies and species the number of different species included.

Table S4: The subset meta‐analytical models evaluating the effects of light pollution on circadian clock, and the influence of experimental conditions: light intensities, days since artificial light at night (ALAN) exposure, time of day, and tissue types. ‘Model index’ contains a series of sequential numbers from 11 to 15, linking the methods to the corresponding results. ‘Data’ refers to the effect sizes included in each model index and k represents the number of effect sizes. ‘Moderators’ show the categories included as moderators in each model, ‘Details’ provide a short description per model and ‘Output section’ show where to find the model output.

Table S5: The effects of artificial light at night (ALAN) on avian circadian clock functioning. We tested these effects in a subset including k = 52 effect sizes reported in two different studies. We conducted a phylogenetic multi‐level meta‐analysis with an intercept‐only model, then, we extended this model by including light intensity as a moderator, and we ran an additional model with time of day as a moderator. The results suggest that ALAN significantly affects circadian clock functioning which increases with light intensity, and these effects are significant at night when compared to the day measure. The table shows estimates for each model in separate rows. Bold estimates indicate confidence intervals (CIs) that did not overlap zero.

Table S6:. The meta‐models testing the influence of experimental regimes (light intensity, days since ALAN exposure, time of day and tissue type) on the avian functional traits' responses to artificial light at night. We assessed subsets including effect sizes that were measured at different experimental conditions and ran a model per each functional trait. The effects of light intensity were also assessed in two subsets, one including all the effects sizes from which light intensity was reported and, and another including effect sizes measured at different light intensities. The tissue‐type subset included effect sizes measured in eight brain regions associated with learning, all grouped as ‘forebrain regions’. The table shows model estimates for each functional trait with bold estimates indicating confidence intervals (CIs) that did not overlap zero. The number of effect sizes k, studies n and species are reported for each level.

Table S7: Studies excluded during each screening phase following the PRISMA flowchart (Figure S1). The research yielded 1507 records (1480 identified through our keyword search and 27 from citation search). After removing 379 duplicates, 1128 unique studies remained and were screened based on title and abstract screening (phase 1). A full‐text screening (phase 2) was performed in 180 records from which 36 studies were included in the meta‐analysis. The table lists the reasons for exclusion and the number of studies excluded in each phase.

Table S8: The effect sizes included in the meta‐analysis on the effects of artificial light at night on avian functional traits that were sign inverted. Since effect sizes in the original studies used a variety of metrics, sometimes reflecting the same biological process in inverse ways, we adjusted the sign of effect sizes such that positive values consistently reflect an increase in the focal functional trait. The table shows the effect sizes that were multiplied by −1 to align metric directionality, relative to each functional trait underpinning physiology (red), behaviour (blue) and life history traits (violet). Effect sizes are accompanied by a ‘Functional note’ as well as a ‘Rationale for inversion’ explaining the reason of sign inversion and the ‘Study’ that reported such effect sizes. The table summarises 86 out of 675 effect sizes from ALAN–DARK paired comparisons representing our dataset.

Table S9: The meta‐analytic mixed‐effect model showing a significant influence of habitat density on the impact of artificial light at night on avian functional traits, for studies performed in the wild. The table shows estimates and confidence intervals (CIs) for each category of the species‐specific trait investigated. The model included: migration and habitat density as moderators. For each level, k represents the number of effect sizes, n the number of studies and species the number of different species included.

Figure S1: Number of studies included in the meta‐analysis on the effects of artificial light at night on avian functional traits. The figure shows the results for each search phase following Preferred Reporting Items for Systematic Reviews and Meta‐Analysis (PRISMA) guidelines. The systematic research of published literature was performed using three different databases: Scopus, Web of Science (All databases) and PubMed, and 27 extra studies were identified and included from citation search.

Figure S2: The effects of artificial light at night (ALAN) on avian circadian clock. (A) The intercept model showing significant and negative ALAN effects on the overall effect sizes. (B) Circadian clock functioning decreased with higher light intensities. The plot show Hedges's g effect sizes (y‐axis) and light intensity (x‐axis), with a solid line representing the model estimate, its 95% confidence intervals (dashed lines) and 95% prediction intervals (dotted lines). (C) ALAN significantly decreased circadian clock functioning at night with day (yellow) and night (grey) time points displayed in the y‐axis. A and C oRchard plots show Hedges's g estimates, its 95% CIs (thick whisker) and 95% prediction intervals (thin whisker). For all plots, positive and negative estimates indicate positive and negative ALAN effects. Asterisks show the significance level (*p < 0.05, **p < 0.01 and ***p < 0.001), k represents the number of effect sizes and the number of independent studies (in brackets) per level. Outputs of full model 11 is shown in Table S5.

Figure S3:. Results from models testing the impact of light intensity on avian function traits. (A) Model including all effect sizes from studies that reported a light intensity measures. Separate subset models were run for each functional trait reporting effects at different light intensities: (B) ageing, (C) immunity, (D) metabolic rate, (E) neuronal cognition, (F) reproductive maturation, (G) sleep regulation, (H) activity offset, (I) activity onset, (J) level of activity, (K) nocturnal activity and (L) body mass. The solid line represents the model estimate, with 95% confidence intervals shown as dashed lines and 95% prediction intervals as dotted lines. The asterisks show the significance level (**p < 0.01 and ***p < 0.001) and k represents the number of effect sizes for each level, while the number of independent studies is shown in brackets. Outputs of full model 12 are shown in Table S6.

Figure S4: Results from model 13 testing the impact of days after artificial light at night (ALAN) exposure on avian functional traits: (A) immunity, (B) reproductive maturation, (C) reproductive success, (D) activity offset and (E) activity onset. We ran a separate model for each functional trait that included repeated effect sizes measured at different days after ALAN. Longer time since exposure to ALAN (more days) significantly increased reproductive success, while the remaining functional traits were not affected. The bubble plots show Hedges's g effect sizes on the y‐axis and days after ALAN exposure on the x‐axis. The solid line represents the model estimate, with 95% confidence intervals shown as dashed lines and 95% prediction intervals as dotted lines. The asterisks *** show the significance level p < 0.001 and k represents the number of effect sizes for each level, while the number of independent studies is shown in brackets. Full model 13 outputs are shown in Table S6.

Figure S5: Impact of artificial light at night on avian functional traits for day and night time points plotted for: (A) ageing, (B) immunity, (C) metabolic rate, (D) neuronal cognition, (E) reproductive maturation, (F) sleep regulation, (G) foraging effort and (H) level of activity. The orchard plots show model estimates for Hedges's g (x‐axis) along with its 95% CIs (thick whisker) and 95% prediction intervals (thin whisker). Positive and negative estimates indicate positive and negative effects of ALAN, respectively. Day (yellow) and night (grey) time points are displayed in the y‐axis. The asterisks show the significance level (*p < 0.05, **p < 0.01 and ***p < 0.001) and k represents the number of effect sizes for each level, while the number of independent studies is shown in brackets. Full model 14 outputs are shown in Table S6.

Figure S6: Tissue‐specific effects of artificial light at night (ALAN) on avian functional traits. Each plot represents a model for a specific functional trait: (A) ageing, (B) immunity, (C) metabolic rate and (D) neuronal cognition. ALAN significantly accelerated ageing in the liver and reduced immunity in hypothalamus, while the rest of functional traits did not show any significant effects. The orchard plots show model estimates for Hedges's g (x‐axis) along with its 95% CIs (thick whisker) and 95% prediction intervals (thin whisker). Positive and negative estimates indicate positive and negative effects of ALAN, respectively. The asterisks show the significance level (**p < 0.01) and k represents the number of effect sizes for each level, while the number of independent studies is shown in brackets. Full model 15 outputs are shown in Table S6.

Figure S7: Results of Model 3 testing for publication bias on effect sizes in the overall model. (A) Significant effects in the small‐study effects model indicate that studies with smaller sample sizes show larger treatment effects than larger studies. The study sample size was transformed using the square root of the inverse of the sample size (x‐axis) and plotted against effect size (y‐axis). (B) No significant effects were detected in the time‐lag bias model testing whether statistically significant effects are published faster than smaller ones. Mean‐centred publication year was included as a moderator (x‐axis). In both bubble plots, the solid line represents the model estimate, dashed lines the 95% confidence intervals, and dotted lines the 95% prediction intervals. Asterisks show significance (*p < 0.05), and k indicates the number of effect sizes per level, while the number of independent studies is shown in brackets.

Figure S8: The effects of artificial light at night (ALAN), tested across wild studies, appeared stronger on avian functional traits of open habitats, while migratory behaviour had no significant influence. Results from the meta‐analytic models including (A) migration, and (B) habitat density. The orchard plots show model estimates for Hedges's g (x‐axis) along with its 95% Confidence Intervals (CIs, whiskers), and the categories of each interspecific trait in y‐axis. Positive and negative estimates assumed a positive or negative effect of ALAN, respectively. The asterisk shows the significance level (*p < 0.05) and k represents the number of effect sizes for each level, while the number of independent studies is shown in brackets. Full model 8b outputs are shown in Table S9.

Diaz‐Palma, S. , Capilla‐Lasheras P., Dominoni D., Cichoń M., and Sudyka J.. 2026. “Artificial Light at Night Consistently Impacts Avian Physiology and Behaviour: A Meta‐Analysis.” Ecology Letters 29, no. 4: e70382. 10.1111/ele.70382.

Data Availability Statement

All R scripts and datasets needed to reproduce the analyses presented in this paper are available at: https://doi.org/10.5281/zenodo.18135290.