Birds that breed exclusively on islands have smaller clutches

ABSTRACT The “island syndrome” refers to similarity in the biology of island organisms, but its generality is questionable, as the scope of species and traits examined are often limited. Here, I show that birds breeding exclusively on islands (breeding island endemics) evolved smaller clutches, using a dataset of 4,530 bird species. Using an inclusive definition of a breeding island endemic, which also encompasses migratory species and seabirds, I examine the evolution of clutch sizes in island breeding species using phylogenetic generalized linear models. Across disparate phylogenetic hypotheses, and after accounting for biological and geographical co-variables, I show that breeding island endemic landbirds (470 species) evolved smaller clutch sizes than continental breeding species (3,818 species). I show that the evolution of clutch size follows the expectations of the island syndrome, as among breeding island endemic landbirds there is a positive relationship between clutch size and breeding range area. Finally, I reinforce the view that the island syndrome is a general pattern in birds, spanning diverse phylogenetic and ecological groups, by showing that in a seabird-only dataset (242 species), breeding island endemic seabirds show evolution of smaller clutch sizes. In a model of the full dataset of both landbirds and seabirds (4,530 species) there was no evidence of an interaction of being a seabird with breeding island endemicity, showing that seabirds and landbirds respond in the same way. This study, using more than 40% of all bird species, provides the first evidence of a general evolutionary response in a life history trait, clearly showing the island syndrome as a general evolutionary tendency associated with island environments. LAY SUMMARY The island syndrome is a phenomenon of similarity in morphology, behavior, and life history between island organisms. Even among birds, its extent is not fully studied; it is not clear which traits (particularly behavioral and life history traits) are involved, and how generally it applies across the whole diversity of birds (more than 10,000 species). Using a dataset of 4,530 species with established molecular phylogenies, I examined whether birds that nest exclusively on islands consistently differ from continental-breeding species in terms of a key life history trait: clutch size. I found that breeding island endemic landbirds have smaller clutches. Among island breeding endemic landbirds, breeding range area was positively correlated with clutch size, as expected in the island syndrome. Seabirds show the same evolutionary response in clutch size as landbirds. This study shows that a life history trait is part of the island syndrome across a representative sample of the avian phylogeny, and shows that seabirds also exhibit the island syndrome. RESUMEN El “síndrome de la isla” se refiere a la similitud en la biología de los organismos de las islas, pero su generalidad es cuestionable, ya que la amplitud de especies y rasgos examinados a menudo es limitada. Aquí, muestro que las aves que se reproducen exclusivamente en islas (reproductoras endémicas de islas) evolucionaron nidadas más pequeñas, utilizando un set de datos de 4.530 especies de aves. Utilizando una definición inclusiva de reproductoras endémicas de islas, que también abarca especies migratorias y aves marinas, examino la evolución de los tamaños de nidada en especies reproductoras de islas mediante modelos lineales generalizados filogenéticos. A través de diversas hipótesis filogenéticas, y después de tener en cuenta las covariables biológicas y geográficas, demuestro que las aves terrestres reproductoras endémicas de islas (470 especies) evolucionaron tamaños de nidada más pequeños que las especies continentales (3.818 especies). Demuestro que la evolución del tamaño de la nidada sigue las expectativas del síndrome de la isla, ya que entre las aves terrestres reproductoras endémicas de islas hay una relación positiva entre el tamaño de la nidada y el área de reproducción. Finalmente, refuerzo la idea de que el síndrome de la isla es un patrón general en las aves, abarcando diversos grupos filogenéticos y ecológicos, al mostrar que en un set de datos exclusivo de aves marinas (242 especies), las aves marinas reproductoras endémicas de islas muestran la evolución de tamaños de nidada más pequeños. En un modelo del set completo de datos de aves terrestres y marinas (4.530 especies), no hubo evidencia de una interacción entre ser un ave marina y ser un ave reproductora endémica de islas, mostrando que las aves marinas y las aves terrestres responden de la misma manera. Este estudio, que utiliza más del 40% de todas las especies de aves, proporciona la primera evidencia de una respuesta evolutiva general en un rasgo de la historia de vida, mostrando claramente el síndrome de la isla como una tendencia evolutiva general asociada con ambientes isleños.


LAY SUMMARY
• The island syndrome is a phenomenon of similarity in morphology, behavior, and life history between island organisms.• Even among birds, its extent is not fully studied; it is not clear which traits (particularly behavioral and life history traits) are involved, and how generally it applies across the whole diversity of birds (more than 10,000 species).• Using a dataset of 4,530 species with established molecular phylogenies, I examined whether birds that nest exclusively on islands consistently differ from continental-breeding species in terms of a key life history trait: clutch size.• I found that breeding island endemic landbirds have smaller clutches.Among island breeding endemic landbirds, breeding range area was positively correlated with clutch size, as expected in the island syndrome.Seabirds show the same evolutionary response in clutch size as landbirds.• This study shows that a life history trait is part of the island syndrome across a representative sample of the avian phylogeny, and shows that seabirds also exhibit the island syndrome.

INTRODUCTION
Geography determines the distribution of species diversity and trait occurrence on Earth (Violle et al. 2014), generating patterns of biological variation following basic features like latitude or insularity (Willig et al. 2003, Whittaker et al. 2023).On islands, area limitation and isolation result in decreased species diversity, and communities with vacant niches (MacArthur andWilson 1967, Whittaker et al. 2023).These island communities are thought to have relaxed predation and inter-specific competition, and higher population densities compared to mainland counterparts (MacArthur andWilson 1967, Whittaker et al. 2023).As these conditions are repeated across the world's islands, they are hypothesized to result in similarities in morphology, behavior, and life history between island-living organisms-"the island syndrome" (Adler and Levins 1994, Lomolino et al. 2013, Jezierski et al. 2023).The island syndrome has been shown to occur in one form or another in all major groups of vertebrates (Benítez-López et al. 2021) and in plants (Burns 2019).Among those, birds are prime island colonizers, with nearly 2,000 out of > 10,000 species being island endemics (Tershy et al. 2015).They exhibit some of the best examples of the island syndrome, including a tendency towards "medium" body size (Clegg andOwens 2002, Benítez-López et al. 2021), a shift towards less flight-reliant locomotion (Wright et al. 2016), or a complete loss of flight (Sayol et al. 2020).However, even in birds, the generality of the island syndrome, in terms of the traits involved and its taxonomic scope (i.e., number of species; clades involved), remains unclear (Meiri et al. 2008).Many behavioral and life history traits have only been partially examined, and phylogenetic comparative methods have seen limited use (Jezierski et al. 2023).These significant omissions preclude a proper understanding of the pattern of the island syndrome, and consequently hinder the study of its evolution (Whittaker et al. 2023) and its contribution to the threats experienced by island species today (Matthews et al. 2022).
Life histories are a key area of study in the island syndrome literature, due to their partial treatment to date (Jezierski et al. 2023), and their importance to individual fitness (Stearns 1992).Previous studies have shown that birds on islands have smaller clutch sizes (Covas 2012), slower growth (Sandvig et al. 2019, Cooney et al. 2020), and higher survival (Beauchamp 2021(Beauchamp , 2022)).However, due to data limitations, many of these studies have examined < 5% of total bird diversity, which cannot be used to ascertain the island syndrome as a global pattern.Only some of them have used phylogenetic trees in their statistical models (Sandvig et al. 2019, Cooney et al. 2020), a problem repeated in other studies of the island syndrome (Jezierski et al. 2023, Whittaker et al. 2023).Furthermore, many studies on the island syndrome have excluded seabirds (e.g., shearwaters, auks or penguins).The reasoning for this exclusion is the unique marine ecology of these birds, which decouples most of their adult life from selection on islands (Clegg and Owens 2002), and makes it difficult to define an island endemic, despite hundreds of seabird species breeding exclusively on islands.The contemporary accessibility and scope of phylogenetic hypotheses (Jetz et al. 2012), as well as extensive data on bird biology (Billerman et al. 2020), makes it possible to examine the evolution of life history traits on islands across hundreds of independent island colonizations, while explicitly including phylogenetic hypotheses.If the island syndrome manifests due to evolutionary change when a species becomes an island endemic, then we should observe a significant association between life history traits and island endemicity.The island syndrome is thought to be exacerbated when species inhabit smaller and/ or more isolated islands, as they are the most limited in terms of species diversity (Lomolino et al. 2012).Therefore, if the observed evolution in life history traits is indeed due to conditions on islands, we should observe evolutionary associations of traits with breeding area in island species, as island area has been shown to affect e.g., body size (Benítez-López et al. 2021).And finally, we can expect that aspects of life histories that take place exclusively on islands will be affected by island conditions.Therefore, by focusing on reproductive life histories of birds on islands, we can create a more inclusive definition of a breeding island endemic-a bird that breeds exclusively on islands, which can include seabirds and migratory species in island syndrome studies.
One such reproductive trait is clutch size, which is a wellunderstood aspect of life history, and is a very direct measure of fecundity (Lack 1947(Lack , 1948;;Bennett and Owens 2002).Bird clutch sizes have been studied extensively through professional and citizen science efforts (Crick et al. 2003), and are one of the best characterized bird life history traits (Jetz et al. 2008).The availability of data on clutch size allows for studies of geographical variation globally, while knowledge of clutch size evolution allows us to make inferences about drivers, and follow up with experimental studies.Islands are thought to generate conditions for "slow" life histories (Covas 2012, Beauchamp 2021), due to lower predation (MacArthur and Wilson 1967) and higher population densities (MacArthur et al. 1972).This results in high life-time survival, and a need to maintain competitive ability, favoring maximizing individual quality rather than fast reproduction (Stearns 1992).In birds, "slow" life histories are characterized by smaller clutch sizes, higher overall survival, and longer life (Robinson et al. 2010).Clutch size and related life history traits vary markedly with multiple aspects of avian biology and distribution, beyond being a breeding island endemic.Larger species of birds tend towards slower life histories, with smaller clutch sizes (Saether 1987).Mode of development (altricial or precocial) is also a significant driver of clutch size, with precocial species having larger clutch sizes, presumably due to their lessintensive parental care allowing for more offspring (Jetz et al. 2008).Finally, migratory species tend to have larger clutches (Jetz et al. 2008), which might be explained by Ashmole's hypothesis that species breeding in seasonal environments (most migrants) will have larger clutches (Lundblad and Conway 2021).This hypothesis has also been invoked to explain the best-known pattern of macro-ecological variation in clutch size, which is the increase in clutch size with latitude (Lack 1948, Jetz et al. 2008).While other explanations could result in lowering clutch sizes towards the equator (Martin 2015), latitude may impact island adaptations, as has been shown in previous studies of clutch size in island birds (Covas 2012).Namely, species living on temperate islands have smaller clutches than expected for a given latitude, which was suggested to stem from lower seasonality on temperate islands (Covas 2012).All these co-variables can confound the patterns of clutch size evolution (Jetz et al. 2008, Covas 2012), and in themselves are important intrinsic factors that determine clutch sizes.If the island syndrome is a general pattern of slower life histories on islands, then clutch sizes should be lower among island breeding endemics, irrespective of these other factors.
Here, I use estimates of avian clutch sizes for 4,530 species (~40% of all bird species) to explore, for the first time, the evolution of a reproductive life history trait on islands across a significant chunk of global bird diversity, using phylogenetic generalized linear modelling.I explore if breeding exclusively on islands results in changes to avian clutch sizes, and how clutch size evolution on islands is mediated by the known biological and latitudinal factors (Jetz et al. 2008, Robinson et al. 2010, Covas 2012).In line with the expectations of the island syndrome, I hypothesize that birds breeding exclusively on islands have evolved smaller clutches than continental-breeding species.To further support the role of island conditions in the observed evolutionary patterns, I explore whether the total area of the breeding range of breeding island endemics impacts the evolution of their clutch sizes.Finally, I examine whether the clutch size evolution of seabirds follows the patterns expected in the island syndrome, and whether seabirds show the same direction of evolutionary change in clutch sizes as landbirds.

Data Collection
Data was collected only on species included in the phylogenies of Jetz et al. (2012) to use explicit phylogenetic hypotheses in explaining clutch size evolution.The two available phylogenies are based on (1) genetic data alone (6,670 species) or on (2) additional inference including subjective decisions (9,993 species).I chose to use the phylogeny based on genetic data alone with the Hackett backbone (Hackett et al. 2008), because it provides a purely molecular and more repeatable phylogenetic hypothesis, used in macro-evolutionary studies of bird traits (e.g., Pigot et al. 2020).The data I collected excludes brood parasites such as cuckoos, because their clutch sizes are not driven by the same trade-offs described above for nesting birds (Payne 1974).I defined breeding island endemic species as a binary variable (yes/no), as birds that exclusively breed on islands, but which may occur outside of islands in different parts of the year (e.g., seabirds).This definition is not commonly used in other studies of the island syndrome, but allows the inclusion of seabirds and migratory species, capturing all species whose breeding attempts are affected by island conditions.Data was collected from Birds of the World (Billerman et al. 2020) between October 2020 and April 2021.For all species in the dataset, I collected the following metadata: species name under the Clements naming convention (Clements et al. 2019); whether a species is a breeding island endemic; minimum and maximum clutch size; and minimum and maximum body mass.If a single value was provided for either clutch size or body mass, it was used as both the minimum and maximum value (including averages).To match the clutch size dataset to both distributional and phylogenetic data used in the analysis, I translated the names of the Clements taxonomy used by Birds of the World to the taxonomy used in Jetz et al. (2012) and BirdLife taxonomy (del Hoyo 2020), using Avibase (Lepage et al. 2014).Additional life history data was collected at family level using Birds of the World, namely development mode (precocial vs altricial), following previous studies on clutch sizes (Jetz et al. 2008).To classify birds as land or seabirds, I classified every bird species that consistently uses marine habitats (coastal, offshore, or pelagic) throughout at least part of its life cycle as a seabird, using Birds of the World and Seabirds: An Identification Guide (Harrison 1985).
All data handling was performed in R v4.2.3 (R Core Team 2023) using the package tidyverse v2.0.0 (Wickham et al. 2019).Geographical data was handled using package sf v1.0-14 (Pebesma 2018).Plotting was performed using ggplot2 v3.4.4 (Wickham and Sievert 2016), gghalves v0.1.4(Tiedemann 2022), and ggdist v3.3.1 (Kay 2023).I used BirdLife distribution maps to obtain data on species' geographic distribution (latitude and area) (BirdLife International and Handbook of the Birds of the World 2020).Because the focus is on a breeding trait, I used BirdLife maps for (1) breeding ranges and (2) year-round ranges, only.The breeding ranges were additionally filtered to include only extant and native distributions.The filtered BirdLife ranges were used to also classify bird species as migratory or not.If a bird species was presented in the BirdLife dataset as having only a "year-round" distribution, it was classified as resident.If it had a separate "breeding-only" distribution, it was migratory, even if the bird was considered resident throughout parts of its range.BirdLife maps may include marine distributions for birds.I excluded these from area and latitude calculations, by intersecting BirdLife maps with land shapefiles from Natural Earth (https://www.naturalearthdata.com/downloads/50mphysical-vectors/50m-land/) at 1:50,000,000 resolution.To determine latitude associated with a given species, I calculated the centroid of each species' distribution using st_centroid() and extracted its latitude.Area was calculated using function st_area().

Modelling Framework
To investigate the impact of geography on clutch size, I used phylogenetic generalized least squares models (Symonds and Blomberg 2014) applied using the R packages nlme v3.1-162 (Pinheiro et al. 2017), ape v5.7-1 (Paradis et al. 2004), and phylolm v2.6.2 (Ho and Ané 2014).Using minimum and maximum clutch sizes would focus on extremes, with maximum clutch sizes being heavily determined by sampling effort (same applies to body mass).Therefore, I calculated the geometric means of both clutch size and body mass to use in modelling.Outliers in clutch size had to be removed, as otherwise the models would break the expectations of normality as examined using the qqnorm() and qqline() functions in the base stats package in R. Manual removing of outliers would increase heterogeneity in the dataset, and potentially introduce additional bias.Therefore, outliers were filtered out of the dataset by removing all species with clutch sizes larger than Q75 + 1.5IQR and smaller than Q25-1.5IQR.Full data for the analysis was available for 4,773 species out of the 6,670 included in the genetic-based phylogenetic data.After removing outliers (243 species), the dataset consisted of 4,530 species (land-and-sea dataset).Of those, 242 were seabirds (seabird-only dataset), and 4,288 were landbirds (landbirdonly dataset).The landbird-only dataset is used for the main analysis, and the analysis of the impact of breeding range area in breeding island endemics (subsampled to 470 species).The seabird-only dataset is used to study whether seabirds show different evolution of clutch sizes on islands.The complete dataset is used to model if there is an interaction between being a seabird and a breeding island endemic.If such an interaction was significant, it would imply that seabirds respond to island environments differently than landbirds in terms of clutch size evolution.
All models were tested for collinearity of explanatory variables using the function vif of package car v3.1-2 (Fox et al. 2012), with 2.5 as the cut-off factor (Zuur et al. 2010).All variables in every model had variance inflation factors < 2. Every linear model was additionally tested for best fit of evolutionary models, comparing Brownian motion; Ornstein-Uhlenbeck models with fixed and random roots (Ho and Ané 2014); Pagel's lambda, kappa and delta (Pagel 1999); and Brownian motion with trend (Ho and Ané 2014) as implemented in phylolm.The performance of evolutionary models was evaluated using the function model.selfrom package MuMIn v1.47.5 (Bartoń 2015) using Akaike Information Criterion corrected for small sample size (AIC c ). Best model is reported in the results, with associated ΔAIC c (AIC c of the second-best model-AIC c of the best model) value to the nearest model.To examine model diagnostics, each model was first run in nlme with a correlation matrix defined by corrPagel() in ape.This approach was used for ease of access to model diagnostics, but it required reorganizing data to match the phylogenetic tree.This was performed using the function comparative.dataas implemented in package caper v1.0.3 (Orme et al. 2018).All models were deemed suitable, and from this point phylolm was used for speed of calculation (both methods arrive at identical results).The trees provided by Jetz et al. (2012) are a posterior sample of thousands of equally likely trees.To account for variation in the posterior sample of phylogenies, I ran each phylogenetic generalized linear model 100 times using different phylogenetic trees from Jetz et al. ( 2012) using a custom approach based on the package sensiPhy (Paterno et al. 2018), collecting all model estimates, errors, p-values, and R 2 values.These methods were applied to all the models described below.

Hypothesis Testing
Every hypothesis was tested using phylogenetic generalized least squares, as described above.To examine whether breeding island endemic species evolved clutch sizes distinct from their continental relatives, the following model was fitted to the dataset of all landbirds (4,288 species): clutch size ~ breeding island endemicity*absolute latitude + developmental mode + migratory status + body mass.Developmental mode, migratory status and body mass were used, as they are known to impact clutch sizes and need to be controlled for (Jetz et al. 2008, Covas 2012).To examine whether breeding island species follow the expectation of more exacerbated adaptations in species inhabiting smaller areas, I used the breeding island endemic-only landbird dataset (470 species) and fitted the following model: clutch size ~ log 10 of the area of breeding range + absolute latitude + developmental mode + migratory status + body mass.In order to test if seabirds follow the expectations of the island syndrome I tested two models: (1) model on seabird-only dataset (242 species): clutch size ~ breeding island endemicity*absolute latitude + developmental mode + migratory status + body mass to test if island and continental breeding seabirds differ; (2) model on seabird + landbird dataset (4,530 species) to test if land-and seabirds differ in their response to being a breeding island endemic: clutch size ~ breeding island endemicity*seabird/ landbird + latitude + developmental mode + migratory status + body mass.All model results are reported in a form of median [5th quantile, 95th quantile].
To investigate the impact of breeding range area on clutch size within the breeding island endemics, Pagel's lambda was again the best evolutionary model (ΔAIC c = 16.24).Among breeding island endemic landbirds (n = 470 species), I found breeding range area to have a significant positive association with clutch size (β = 0.093 [0.089, 0.095], p-values = <0.001; Figure 3).In the dataset of breeding island endemics, latitude was still a significant positive predictor of clutch size (β = 0.024 [0.023, 0.025], p-values = <0.001),but developmental mode, migratory status and mass were not significant (Supplementary Material Table 2).The model explained less variation in breeding island endemic species only (R 2 = 0.154 [0.146, 0.161]), and phylogenetic signal was even higher than in the landbird dataset (λ = 0.901 [0.883, 0.920]).
In a dataset of both land-and seabirds (n = 4,530 species), Pagel's lambda remained the best evolutionary model (ΔAIC c = 627.98).The interaction term between breeding island endemicity and being a seabird was non-significant (β = 0.113 [0.093, 0.131], p-values = 0.315 [0.248, 0.415]).Meanwhile, being a seabird results in a lower clutch size across all phylogenies (β= -0.493 [-0.532, -0.454], p-values ≤0.001).The addition of seabirds to the dataset resulted in development mode no longer being significant, whereas all other variables remained qualitatively the same after including seabirds in the model (Supplementary Material Table 4).

DISCUSSION
The results of this study confirm that birds breeding exclusively on islands have smaller clutches, in accordance with the expectations of reproductive life history forming a part of the island syndrome.Using one of the largest samples of species to study the island syndrome to date, the results reinforce the previous findings on bird reproduction (Covas 2012), and the FIGURE 2. Clutch sizes increase with latitude, but with no interaction of being a breeding island endemic or a continental breeder (solid linecontinental species; dashed line-breeding island endemic species).The lines were parametrized using an example phylogenetic tree, the 95% confidence interval is shown by gray shading (tiny and thus not visible for the continental line).Note: Horizontal point jitter added for clarity.Landbird-only dataset (n = 4,288 species). of slower avian life history on islands (Covas 2012, Sandvig et al. 2019, Cooney et al. 2020, Beauchamp 2021, 2022).I found that lower clutch sizes evolved in breeding island endemics, even if latitude and other aspects of life history are considered.Using an explicitly phylogenetic modelling framework implies that the observed differences co-occur with becoming a breeding island endemic along bird evolutionary history.This means that selection caused by island environments is likely acting on breeding island endemics, and the observed patterns of the island syndrome at least partially stem from evolutionary change and not only, for example, filter effects associated with successful island colonization (Lomolino et al. 2012).Evolutionary change towards an element of slower life histories reinforces the notion that it might be an adaptation to low-predation, high-conspecific competition environments (MacArthur and Wilson 1967, Lomolino et al. 2012, Whittaker et al. 2023), which are thought to characterize islands, provided such conditions occur (which could be examined in field studies).
Evolution of smaller clutch sizes on islands was recovered in every model, even if other biological and geographical co-variables were considered.This yielded the somewhat surprising result of body mass not being a significant predictor of clutch size, contrasting with some previous work (Saether 1987, Jetz et al. 2008), but in agreement with others (Martin et al. 2006).However, some of these previous works have considered limited geographical and taxonomic samples of birds (Saether 1987, Martin et al. 2006), and none of them considered phylogenetic relationships in their models, which may explain the discrepancy in results.Other aspects of life history such as developmental mode and migratory status correlated with clutch size as expected, in line with the results of previous studies (Jetz et al. 2008).
My results corroborate the pattern of increased clutch size with absolute latitude (Lack 1948, Jetz et al. 2008).However, I did not recover the previously reported interaction of latitude with island endemicity, where temperate species lay smaller clutches on islands for a given latitude (Covas 2012).There are significant differences in design between that study and the one reported here, and likely reasons for the different results include significantly larger sample size and explicit phylogenetic modelling in this study.Covas (2012) invoked stability of climate on islands as a potential driver of the island syndrome and argued that the interaction may be driven by lower seasonality on temperate islands.The discrepancy between our two studies suggests that this could be explored using the dataset presented here in conjunction with environmental data.A further idea to explore is that latitude often acts as a catch-all variable that strongly covaries with productivity and strength of biotic interactions.For birds, biotic interactions are thought to increase with lower latitude, for example, through higher predation risk (McKinnon et al. 2010, Freeman et al. 2020).Islands should lower the slope of that gradient due to a significant drop in the risk of predation, and thus also result in an interaction between latitude and island endemicity, although of the opposite direction to the one found by Covas (2012).
To further reinforce the fact that the unique selective pressures on islands lead to the evolution of traits encompassed by the pattern of the island syndrome, I showed that breeding island endemics inhabiting smaller breeding areas evolved even smaller clutches.Island area is a significant determinant of community structure, with smaller islands tending to be more disharmonic and losing functional diversity (MacArthur andWilson 1967, Whittaker et al. 2023).As empty niches and low predation are the proposed drivers of the island syndrome (Lomolino et al. 2012, Ponti et al. 2023), it should be the strongest on the smallest islands.It is indeed the case across some measurements (e.g., in body sizes of island vertebrates; Benítez-López et al. 2021), but island area has not always been recovered as an important predictor (Schwarz et al. 2020).The result of this study shows that area impacts clutch size evolution.However, due to differences in study design, there are some limitations to the interpretation.While other studies have performed meta-analyses of studies from particular islands, where a measurement of a trait can be linked to an exact island (Benítez-López et al. 2021), my study examines a species-wide mean.While many island-endemic species or breeding island endemics inhabit individual islands, there are species that inhabit entire archipelagos and move between islands (e.g., the Nicobar Pigeon [Caloenas nicobarica]; Billerman et al. 2020), yet still undoubtedly are limited to islands.Therefore, the studied value of clutch size cannot be easily linked to an individual island, and the analysis is thus not identical to previous studies.However, this still strongly suggests that insularity drives trait evolution, and can be further investigated in a meta-analytic framework in which island populations are studied.
This study has, for the first time, explicitly examined if seabirds also show patterns of trait evolution consistent with the island syndrome.It was possible due to the use of a more inclusive definition of being a breeding island endemicwhich many seabird species are, even though their adult life is largely spent at sea.The results of this study show that breeding island endemic seabirds evolved smaller clutches than their continental breeding seabird relatives, just like landbirds do.The impact of breeding island endemicity was nearly double that of the effect the landbird dataset, due to the frequent clutch size of a single egg in island breeding seabirds.Furthermore, by analyzing land and seabirds together, I show that the direction of evolutionary change in clutch size in breeding island endemics is the same between seabirds and landbirds.Even though all seabirds have significantly lower clutch sizes than landbirds, there is no interaction between being a seabird and breeding island endemicity in the model.Therefore, seabirds may be needlessly excluded from studies of the island syndrome (Clegg and Owens 2002), or at least should not be excluded when the studied traits can directly, or indirectly, be affected by selective conditions on islands.Further inclusion of seabirds in island syndrome studies could be particularly fruitful, as many traits of seabirds (slow life history, evolution of unique ecologies) are in general considered parts of the island syndrome (Jezierski et al. 2023, Whittaker et al. 2023).
To conclude, this study is the largest characterization of a reproductive life history trait in the context of the island syndrome, and one of the largest investigations into the island syndrome at all (Benítez-López et al. 2021) in terms of the number of species considered.Its results conclusively show that breeding island endemicity is associated with the evolution of smaller clutch sizes.By comparing different datasets and aspects of evolution on islands (i.e., breeding endemicity, area), I show that the recovered evolutionary patterns in clutch size evolution are consistent with the island syndrome.Furthermore, thanks to the taxonomic scope and inclusion of seabirds, I have been able to demonstrate that the evolutionary change on islands applies across all birds, and across very disparate avian lifestyles, that nevertheless converge on islands during reproduction.

FIGURE 1 .
FIGURE 1. Landbirds breeding exclusively on islands have smaller clutches, an effect significant regardless of phylogenetic hypotheses or other factors.(A) Distributions of geometric means of clutch size between continental breeding species and breeding island endemics.Note: Horizontal point jitter was added for clarity using function geom_ jitter() of ggplot2.Each point corresponds to a single species.Panels (B) and (C) show the results of phylogenetic generalized linear models run across 100 phylogenies obtained from Jetz et al. (2012), with the distribution of (B) model β coefficients with the vertical line at 0; and (C) p-values with the vertical red line at 0.05 to denote the significance threshold.Distributions show 5th to 95th quantiles.Landbird dataset (n = 4,288 species).

FIGURE 3 .
FIGURE 3. Breeding island endemics with smaller breeding areas have smaller clutches.Each point corresponds to a different species.Thick black line represents the line of best fit, with grey shading representing the 95% confidence interval.The line was parametrized using an example phylogenetic tree.Note: Horizontal point jitter added for clarity.Breeding island endemic (landbirds) dataset (n = 470 species).

FIGURE 4 .
FIGURE 4. Seabirds breeding exclusively on islands have lower clutch sizes than continental-breeding relatives.Slabs to the right of boxplots show the distribution of clutch sizes per species in each category and are scaled to the total number of species in a given category.Complete dataset (n = 4,530 species).