Egg mass drives the evolution of bird nest architecture

Abstract

Although most birds build nests, not all of them build nest walls, and the evolutionary advantage conferring this architectural feature remains inconclusive. By integrating macro-evolutionary patterns with individual fitness consequences, we investigated the function of nest walls in protecting eggs and the evolutionary driver behind this construction. We examined two novel hypotheses relating to egg rolling-off risk: (i) heavy-bird hypothesis: heavier birds generate greater nest vibrations during attendance, increasing egg rolling-off risk; and (ii) light-egg hypothesis: lighter eggs are more susceptible to displacement from nest vibrations due to lower friction against the substrate. Phylogenetic comparative analyses across 4030 species revealed that birds with lower body and egg mass were more likely to construct nest walls, suggesting higher rolling-off risk for lighter eggs. Experiments combining behavioural observations and controlled trials with three-dimensional-printed eggs confirmed that lighter eggs were more prone to falling from nests without walls. The evidence across phylogeny and individual-level experiments consistently supports the light-egg hypothesis, suggesting the adaptive function of nest walls in preventing eggs from falling during external disturbance, particularly for lighter eggs. This study demonstrates how integrating macro- and micro-evolutionary approaches can reveal the functional mechanisms underlying correlated evolution between phenotypic traits.

1. Introduction

Growing awareness in evolutionary ecology highlights the need to reconcile micro- and macro-evolutionary frameworks to deepen our understanding of the evolutionary processes that shape biodiversity [1–5]. Macro-evolutionary studies typically examine phenotypic shifts across lineages over extended timescale, providing insights into niche evolution and species divergence [1,6]. In contrast, micro-evolutionary studies examine the mechanisms (e.g. plasticity, drift, mutation, selection) that drive phenotypic variation within species and facilitate local adaptation [1,7,8]. Although these two approaches operate at different temporal and taxonomic scales, they complement each other and their integration can offer new insights. Advances in comparative methods have enabled macro-evolutionary studies to formulate testable hypotheses based on the observed patterns of phenotypic evolution and diversification. Micro-evolutionary studies, in turn, can empirically test the mechanistic bases of these hypotheses by demonstrating how selective forces shape the adaptive fitness of phenotypic traits below the species level. Despite the clear potential for synergy between these approaches, few studies have applied this integrative framework to examine the drivers of phenotypic evolution.

Bird nests represent one of the most diverse phenotypes in the animal kingdom, and its evolution offers tremendous opportunities to study phenotypic evolution across different biological scales. Bird nests are conventionally regarded as structures primarily designed to protect eggs [9–12] and enhance incubation efficiency [13]. Thus, nests and eggs are expected to collectively influence individual reproductive success [9,14–16], although evidence remains mixed, probably due to confounding effects of spatio-temporal variations in habitats, weather or individual quality [13,17]. Comparative studies have greatly contributed to our understanding of bird nest evolution and its adaptive benefit [15,18–20], primarily by examining the evolutionary associations between various nest characteristics (e.g. structure and site [20–22]) and egg traits (e.g. size, shape and pigment [19]) across the bird phylogeny. Along the avian evolution history, bird nest structure has evolved from simple platforms to more complex cup or dome structures with erected walls [20]. Concurrently, bird eggs have shifted from larger to smaller size, mirroring the trend in body size evolution [19]. These macro-evolutionary patterns suggest that nest walls, body mass and egg mass may be evolutionarily associated with each other [19]. However, the functional adaptations linking nest structure to egg mass have received relatively limited empirical attention, as have the mechanistic pathways driving the evolutionary diversification of these phenotypes [11].

Since bird nests and eggs are evolutionarily linked, they can act as selective pressures on each other or be shaped by the same selective forces. Meanwhile, each of them is also subject to distinct evolutionary pressures, further contributing to the intricate dynamics of nest and egg adaptation. Identifying these selective forces and how they act on these evolutionarily associated traits enhances our understanding of phenotypic diversification across the bird phylogeny. Traditional hypotheses on the functional associations between bird nests and eggs [23–27] posit how enclosed nests, compared with open nests, provide greater benefits in thermoregulation [25] or reduced predation pressures for birds [28,29]. Yet, enclosed nest structures impose costs, such as requiring more time and energy to build [14] and increasing the risk of egg damage due to collision in the nest [30]. These contrasting effects of nest structures on egg survival highlight the complexity of nest–egg relationships. However, the multifaceted views potentially obscure the fundamental pattern in their correlated evolution that might otherwise be detected from a simpler perspective.

Eggs falling out of nests directly cause reproductive failure [31,32], and thus, preventing egg displacement could be the most obvious benefit for birds to build nest walls [13]. However, it is surprising that this functional benefit has rarely been examined. If egg retention is indeed one important function of nest walls, we would expect that birds facing a higher risk of egg rolling-off tend to build nests with wall structures. Thus, we propose two novel hypotheses linking birds’ body or egg mass to nest architecture through the egg displacement risk—the heavy-bird hypothesis and the light-egg hypothesis. The heavy-bird hypothesis posits that heavier parent birds generate greater nest vibrations during nest attendance due to increased gravitational force, elevating the risk of eggs rolling out of nests. This hypothesis predicts that eggs produced by heavier birds face higher displacement risks in nests without walls, and consequently, heavier birds are more likely to build nest walls. Alternatively, the light-egg hypothesis focuses on egg properties, suggesting that lighter eggs have lower friction against the nest substrates and reduced rotational inertia (the torque required for rotational acceleration [33,34]). These physical properties make lighter eggs less resistant to nest vibration caused by external forces (e.g. wind [31]) and are more prone to rolling out of nests. Accordingly, this hypothesis predicts that lighter eggs face higher rolling-off risks in nests without walls and that birds producing lighter eggs should exhibit evolutionary tendency towards building nests with walls. Given that heavier birds tend to produced heavier eggs, the two hypotheses represent two mutually exclusive mechanisms underlying nest wall evolution in relation to rolling-off risk.

To test the two competing hypotheses, we employed an integrative framework combining phylogenetic comparative analyses and experimental manipulation. First, we examined the evolutionary associations of nest wall structure with both body mass and egg mass across the bird phylogeny while considering phylogenetic distance among species. In addition to examining absolute egg mass, we also assessed the effect of relative egg mass—a measure indicating how resistant an egg is to the gravitational force generated by parent bird’s body mass during nest attendance. Second, to directly assess the functional benefit of nest wall structure in preventing egg displacement, we implemented behavioural experiments with society finches (Lonchura striata domestica). We compared egg-dropping rates between birds building nests with and without walls across varying levels of body mass. We also manipulated nest wall structure (i.e. with or without walls) and egg mass (i.e. heavier or lighter) to estimate the likelihood of eggs rolling out of nests under each experimental combination. Our results reveal that egg mass serves as a stronger evolutionary driver of nest wall development than body mass when considering the risks of eggs rolling off. This study provides novel insights into the selective pressures shaping avian nest architecture.

2. Material and methods

(a). Phylogenetic analyses

We examined the evolutionary association between bird/egg mass and nest wall structure across the avian phylogeny (figure 1a) using phylogenetic generalized linear regression models. We obtained the information on nest structure from the global database of bird nest traits [35] and selected a total of 4030 species that build platform, cup or domed nests. A platform nest has lining on the bottom, intertwined into a platform, with none or little shielding on the edge. A cup nest is shaped like a cup with an erected, surrounding rim that is not deep enough to shield an adult bird. A domed nest has a domed or sphere shape and allows an adult bird to sit inside without exposing itself. We considered platform nests as the nests without walls; cup and domed nests were categorized into the nests with walls because they feature the nest structure vertically surrounding the base of nests. If a species can build nests both with and without walls, we considered it as the one building nests with walls. This approach reflects that wall construction represents more complex behaviour than building nests without walls, and previous research [20] shows that nests with walls are evolutionarily derived relative to nests without walls. This evolutionary perspective suggests that species capable of wall construction possess the architectural repertoire to build both nest types, while species building only nests without walls may lack this construction capability. Consequently, classifying architecturally versatile species by their most complex nest type better reflects their evolved capacities. To test the robustness of our results to the classification, we repeated the analysis by classifying ambiguous species as ones building nests without walls and obtained generally consistent results (electronic supplementary material, results S1, table S1). Thus, we only reported results based on the primary, more justifiable classification criterion in the main text.

Figure 1.

Sketches of research approaches used in this study. (a) Phylogenetic comparative analyses were used to examine the evolutionary association between bird/egg mass and nest wall structure across the avian phylogeny using phylogenetic generalized linear regression models. (b) Nest construction experiments were used to examine the effect of nest walls on the number of eggs dropped from society finch nests with the weights of parent birds included as covariates. (c) Incubation experiments were used to examine the effects of nest walls and egg mass on the likelihood of egg dropping by manipulating the walls of artificial nests and the weight of artificial eggs in three treatments (T0, T1 and T2) to society finches. Sketches by Hsiang-Ching Chen.

We extracted the data of egg mass from Rotenberry & Balasubramaniam [36] and clutch size from the dataset ‘Amniote’ assembled by Myhrvold et al. [37] and calculated clutch mass as the product of clutch size and egg mass for each species. The primary source of body mass data was AVONET [38], while missing data in this source were substituted by information from the Handbook of the Birds of the World Alive [39]. We used the residuals of a simple linear regression model with the egg/clutch and body masses being the dependent and independent variables, respectively, as the measurement of relative egg/clutch mass for individual bird species. Given the wide range of body, egg and clutch masses across bird species, we transformed the mass data using the natural log for further analysis. We found that the analytic results of egg and clutch mass were broadly consistent (electronic supplementary material, results S2, table S2) and only reported egg mass results in the main text because this variable is directly related to the light-egg hypothesis.

In our analyses, we also incorporated two additional covariates: passerine/non-passerine classification and nest site (ground/non-ground). We included these two factors because passerines typically construct nests with walls and have employed diverse nest attachment and site utilization strategies to establish new nesting niches since their major adaptive radiation [20]. As such, their nest wall construction behaviour may be influenced by evolutionary forces distinct from those operated in non-passerines. We categorized species as either being ground nesters or non-ground nesters based on the information also from the nest trait database [35]. Ground nesters build nests on the ground or underground, and thus, eggs rolling out of the two places are less likely to be broken. Non-ground nesters, on the other hand, construct their nests in trees, non-tree vegetation, cliff/banks, ant/termite nests or on water bodies. Eggs rolling out of these nests would probably lead to egg loss directly (e.g. [31,32]). For the species that exhibit both ground and non-ground nesting behaviours, we classified them as non-ground nesters because they would have necessarily evolved strategies to mitigate the elevated rolling-off risk associated with non-ground nests—without such adaptation, the fitness cost would have precluded the maintenance of both nesting strategies within the same species. This classification approach enabled our analysis to capture the evolutionary adaptations most relevant to nest wall construction in various nest sites. We also conducted models to a dataset containing exclusively non-ground nesters. We found that the main effects of absolute body mass, egg mass and relative egg mass on nest wall existence remained consistent with the results of our full models including both ground and non-ground nesters, though the interaction between passerine status and egg mass differed across models focused on absolute egg mass (electronic supplementary material, results S3, table S3). Thus, we only reported results based on the full models.

We managed the data and performed the analyses in R v. 4.0.5 [40]. We built the phylogenetic models using ‘phyloglm’ function (method = logistic_MPLE, given that presence/absence of nest walls is a binary dependent variable) in the ‘phylolm’ package [41]. We performed Pearson’s correlation analysis to show a strong positive correlation between log-transformed body and egg mass (r = 0.98, p < 0.001). Thus, we analysed the effects of these mass traits independently in different models. Each model consisted of an explanatory variable (i.e. absolute body or egg mass or relative egg mass) and two covariate variables (i.e. passerine/non-passerine and ground nester/non-ground nester), as well as their interactions with the mass variables (i.e. mass and passerine, mass and nest site). We applied the natural log-transformation and standardization to the mass-associated data before building the models.

To consider the possible effect of phylogenetic relationships between species in the analyses, we built a consensus tree by randomly extracting 1000 tree replicates from BirdTree (http://birdtree.org) based on the Hackett constraint. We mapped the information of nest wall and weight-associated data on the phylogeny after generating a majority-rule consensus tree from the 1000 trees with the ‘phytools’ package [42]. Given that phylogenetic comparative analysis can be sensitive to uncertainty in phylogenetic reconstruction and tree topology, we conducted phylogenetic sensitivity analyses for the models using ‘sensiPhy’ v. 0.8.5 [43] to assess the robustness of our results across alternative phylogenetic hypotheses. We estimated the impacts of phylogenetic uncertainty on our model results using the function ‘tree_phyglm’ with 100 alternative phylogenetic trees.

(b). Experiment preparation

To examine the effect of nest walls on preventing eggs from falling off nests, we conducted two experiments using the society finch. The species exhibits great variation in nest structure it builds (electronic supplementary material, figure S1), including platform, cup and dome structure, and thus allows us to have enough broad spectrum to test the functional connection between eggs and nests, without the need for controlling the species effects if using multiple species. They also spontaneously incubate and foster heterospecific clutches, and their parenting behaviour can be less disturbed by investigator inspection [44]. This aspect is particularly advantageous for our study as it allows us to conduct a simulated egg incubation experiment, facilitating the investigation of the complex associations between parent weight, egg mass (using three-dimensional-printed eggs) and nest walls (using both bird-built nests and artificial nests; see details below).

Forty-six pairs of society finches participated in the experiment of nest construction (figure 1b), and 18 females were the subject of the experiment of three-dimensional-printed egg incubation in artificial nests (figure 1c; see below for details). The society finches were purchased from pet shops in Taipei City. Before the beginning of experiments, the birds were habituated in a bird room at 23−25°C, approximately 40% humidity and under a 13 h : 11 h light : dark cycle for at least a month. We collected blood samples from each individual to identify sex and housed groups of males and females separately in wire cages (1 × 1 × 1 m). The birds had ad libitum access to water and food comprising a mixture (by volume) of white millet (75%) and canary seed (25%), with additional shell powder (10 g) and multivitamin (5 g). All of the birds were naive to nest construction before the experiments.

(c). Nest construction experiment and reproductive performance

We randomly assigned the birds into male-female pairs and transferred them to the experimental cages (45.5 × 37.0 × 42.5 cm) in another bird room under a 14 h : 10 h light : dark cycle and with temperature and humidity same as those of the room used for habituation. Each of the cages contained two perches and a horizontally mounted cardboard box (15 × 15 × 3.5 cm), inside which society finches could construct a nest. We provided 10 g of sterile coconut fibres every day as nest materials; the mass of coconut fibres used was recorded each day by deducting the mass of fibres that remained from the previous day.

As nest construction progressed, we considered the nests as completed when the pairs utilized less than 2 g of nest materials for 3 continuous days. Nests were photographed on the day of completion and categorized as either with or without walls. We considered nests built without walls if there were no nest materials surrounding and extended vertically from the basal nest area and vice versa (figure 1b; electronic supplementary material, figure S1). Among the breeding pairs, we recorded the amount of nest materials used by summing up the coconut fibres used each day. Incubation began when the first egg was laid and lasted until the first chick was hatched. We also recorded the number of eggs falling off nests during the incubation period, as well as chicks and eggs remaining in the nest (i.e. clutch size) at the end of the incubation period. Given that the experimental cages were located within the controlled bird room, eggs could potentially fall off nests only when parental birds caused nest vibration upon entering or leaving the nests, or (accidentally) pushed the eggs out.

We analysed the effect of the existence of nest walls on four response variables—(i) the number of eggs falling off nests, (ii) clutch size, (iii) the amount of nest materials used, and (iv) nest construction duration—using R v. 4.0.5. We employed negative binomial regression models for egg loss and clutch size, and simple linear regression models for nest material mass and the duration of nest building. We transformed the values of nest mass using the natural log for meeting the model assumption of having a normal distribution. For each response variable, we first constructed an initial model including the existence of nest walls as a fixed factor. We also included the weight of mother and father birds as the covariates in the model for two reasons. First, there is a close relationship between parent weight and egg mass (see above); second, birds in different weights may potentially cause different magnitude of nest vibration. We then performed the bidirectional model selection and selected the final models based on the Akaike information criterion values, excluding models showing singularity.

(d). Incubation experiment of model eggs in artificial nests

We subsequently conducted an incubation experiment to examine the function of nest walls in securing eggs of different weights in nests by manipulating nest structure and the mass of three-dimensional-printed eggs (ZRapid SLA500 three-dimensional printer, figure 1c). The experiment was conducted in the same laboratory conditions mentioned above. The society finch generally has a clutch size of four to six eggs [45], so we put four three-dimensional-printed eggs with either regular (heavy eggs; 1.10 g per egg) or reduced weight (light eggs; 0.93 g per egg) in two types of artificial nests (i.e. with and without walls); the artificial nests were placed above ground in the experimental cages (electronic supplementary material, figure S2). The basal areas of the two types of nests were the same (diameter = 11.5 cm; [46]), while nests with walls had a depth of 4 cm measured from the outer surface. Model eggs were consistent in size with mass variance lying within 5% of the target mass. We decided the egg mass based on the mean and the 2.5th percentile of real, fresh eggs laid by the society finch (n = 37). We weighed the birds before they undertook the experiment. There were three treatments: nest without a wall and light eggs (T0), nest without a wall and heavy eggs (T1) and nest with a wall and light eggs (T2). We did not implement the treatment of nest with walls and heavy eggs, as it would become a control trial with differences in more than one variable (i.e. egg mass and nest wall) compared with T0, making it difficult to isolate the effects of explanatory variables.

Eighteen female birds participated in the experiment, and each of them experienced all of three treatments. This experimental set-up enabled us to control the effect of nest vibration caused by parent birds with different weight and examine the functional advantages of nest walls in reducing the risk of eggs rolling off. To eliminate the potential effects of incubation experience (i.e. the order of the treatments) on the results, we assigned the birds to the three treatments in one of the systematic orders (i.e. T0-T1-T2; T0-T2-T1; T1-T0-T2; T1-T2-T0; T2-T0−1; T2-T1-T0). The trials began as soon as the focal female bird made contact with the nest; they were completed once there was no egg left on the nest or on day 11, the timing that most eggs produced by the society finch attained the lowest weight (electronic supplementary material, figure S3). We defined a trial as failed if the bird did not touch the nest throughout the trial. The birds that failed two trails consecutively were replaced by a new bird. After a trial was completed, we allowed the birds to rest for at least a week before resuming the next trial. To reduce effects of anthropogenic disturbance on experimental outcomes, we mounted a digital camera (Tin-Wei AW−720CIP) above each experimental cage to monitor the number of eggs remained in the nests (electronic supplementary material, figure S2b). We recorded presence/absence of each focal bird on the nest every 15 minutes as the proxy of duration of nest attendance.

We used an ordinal regression model to examine the effects of nest wall existence and egg mass on the number of eggs falling off nests, which was an ordinal variable ranging from zero to four. We set the types of treatment (T0, T1 and T2) as the explanatory variable and included both the body weight of female birds and how many times each bird was observed in its nest (i.e. nest attendance) as the covariates in the ordinal regression model. This model also estimated odds for different numbers of eggs dropped.

3. Results

(a). Light body and egg mass were associated with construction of nest walls across the avian phylogeny

Our phylogenetic analyses of absolute mass revealed that species lighter in body or egg mass tended to construct nest walls (figure 2, table 1), supporting the light-egg hypothesis, but not the heavy-bird hypothesis. Passerines, which generally have lighter body and egg mass compared with non-passerines, exhibited a higher tendency for nest wall construction even when their body or egg mass became heavier (figure 2). In addition, the associations between body/egg mass and nest wall construction were consistent between ground and non-ground nesters (figure 2, table 1). Such trends were validated by the sensitivity analyses that considered phylogenetic uncertainty (electronic supplementary material, figure S4). In the analysis of relative egg mass (egg mass controlled for body mass), we found no significant effect of this trait on nest wall building tendencies overall (table 1). However, our analyses revealed an interaction between nest site and relative egg mass: non-ground nesters showed a stronger tendency to build nest walls compared with ground nesters, but this tendency diminished as relative egg mass increased (figure 3, table 1).

Figure 2.

The predicted probability of nest wall construction considering values of absolute (a) body and (b) egg mass after the natural log-transformation. We also included two binary factors (i.e. passerine/non-passerine and ground nester/non-ground nester) in each of the models and tested their effects on nest wall construction across the bird phylogeny. The trends based on our model predictions are depicted by solid lines, while the observed data points are represented by dots in the graphs.

Table 1.

Statistical values were derived from models that investigated the evolutionary associations between nest wall construction and the absolute egg mass, absolute body mass and relative egg mass. Each of these models also included two interactive effects (i.e. mass and birds being as a passerine, mass and nest site). We applied the natural log-transformation to the mass data before building the models. Phylogenetic correlation was assessed using alpha values, which provided an estimate of the degree of conservation observed in the examined traits across species. A lower alpha value indicates a higher level of trait conservation. Bold p-values indicate statistical significance.

model	fixed factors	N	estimate	p-value	alpha
wall ~ egg + passerine + site + egg*passerine + egg*site	egg	4030	−2.255	<0.001	0.005
	passerine		1.485	0.080
	site (non-ground)		−0.165	0.601
	egg*passerine		1.682	<0.001
	egg*site (non-ground)		0.292	0.325
wall ~ body + passerine + site + body*passerine + body*site	body	4030	−1.646	<0.001	0.007
	passerine		1.569	0.066
	site (non-ground)		0.384	0.111
	body*passerine		1.563	<0.001
	body*site (non-ground)		−0.196	0.362
wall ~ rel.egg + passerine + site + rel.egg*passerine + rel.egg*site	egg	4030	0.154	0.224	0.005
	passerine		1.975	0.081
	site (non-ground)		1.408	<0.001
	egg*passerine		0.256	0.355
	egg*site (non-ground)		−0.406	0.030

Figure 3.

The predicted probability of nest wall construction considering relative egg mass and its interactions with the two binary factors (i.e. passerine/non-passerine and ground nester/non-ground nester) across the bird phylogeny. The trends based on our model predictions are depicted by solid lines, while the observed data points are represented by dots in the graphs.

(b). Nest walls reduced egg loss in light-weight birds

We categorized nests built by society finch pairs in the nest construction experiment into with-wall (n = 23 pairs) and without-wall groups (n = 18 pairs) and elucidated the functional benefit of nest walls in individual reproductive success. We found strong evidence that nest walls play a crucial role in reducing the rolling-off risk of eggs, particularly in lighter females. Both the presence of nest walls and heavier mothers significantly reduced the number of eggs dropped from nests (nest wall: estimate = −27.930, s.e. = 7.956, Z = −3.511, p < 0.001; mother mass: estimate = −0.772, s.e. = 0.369, Z = −2.093, p = 0.036; figure 4a; electronic supplementary material, table S4). There was a significant effect of the interaction between nest wall and mother body mass on egg loss (estimate = 1.075, s.e. = 0.497, Z = 2.164, p = 0.03; electronic supplementary material, table S4). Eggs produced by lighter mothers in nests without walls faced the highest risk of rolling off, while nest walls effectively mitigated this risk. Conversely, father mass showed no significant effect on the number of eggs dropped (father mass: estimate = −0.456, s.e. = 0.327, Z = −1.404, p = 0.160; electronic supplementary material, table S4). The results provide strong support for the light-egg hypothesis because lighter female birds typically produce lighter eggs that are more susceptible to rolling off nests without structural barriers.

Figure 4.

Effects of the presence of nest walls (and mother weight) on (a) the number of eggs dropped during incubation, (b) clutch size, (c) nest mass and (d) the duration of nest building. Solid lines in (a) illustrate the trends based on our model predictions while dots in graphs illustrate the observed data.

We examined additional reproductive or nesting parameters, but found no significant effect of nest walls on clutch size (estimate = 0.278, s.e. = 0.183, Z = 1.516, p = 0.129), amount of nest materials used (estimate = −0.087, s.e. = 0.143, t = −0.61, p = 0.545) or duration of nest building (estimate = −0.486, s.e. = 0.723, t = −0.672, p = 0.506; figure 4, electronic supplementary material, table S4).

(c). Experimental manipulation confirmed that nest walls and egg mass reduced egg rolling-off risk

Our incubation experiments revealed that both increased egg mass (T1 vs. T0) and the presence of nest walls (T2 vs. T0) significantly reduced the likelihood of eggs rolling off nests (p < 0.01; table 2). These findings confirm the protective effects of both factors identified in the nest construction experiment. Interestingly, the likelihood of eggs rolling off nests increased with nest attendance by parents at a marginally significant level (p = 0.053; table 2), and nest attendance had distinct effects depending on nest structure (p < 0.001; table 2). In the nests without walls, increased nest attendance rapidly elevated the likelihood of heavy eggs rolling off nests (Attendance*T1). In contrast, nest walls mitigated this risk as increased attendance had little impact on the dropping rate of light eggs in nests with walls (Attendance*T2). We found significantly higher odds of fewer eggs drop (p < 0.001 for 1 vs. 2, 2 vs. 3 and 3 vs. 4, but not for 1 vs. 0; table 2), suggesting that eggs rolling out of the nest were not random events.

Table 2.

Results of an ordinal regression model showing the effect of nest walls, egg weight and nest attendance, as well as their interactions, on reducing the number of eggs falling out of nests. Each bird experienced three treatments (n = 18; T0: nests without walls and light eggs, T1: nests without walls and heavy eggs, T2: nests with walls and light eggs) and we provided each of them with four eggs in the same weight in each experimental trial. The intercepts (i.e. x | x + 1, x = 0–3) represent the log of odds of x egg(s) dropped versus x + 1 eggs dropped. Values in the table indicate coefficients, standard errors (s.e.), t values, 2.5 and 97.5% confidence intervals and p-values, respectively. Bold p-values indicate statistical significance.

	coefficient	s.e.	t value	2.5%	97.5%	p-value
T1 (egg mass)	−2.704	0.861	−3.140	−4.392	−1.016	0.002
T2 (nest wall)	−18.474	<0.001	−1.136*106	−18.474	−18.474	<0.001
attendance	6.501	3.363	1.933	−0.091	13.093	0.053
attendance*T1	39.225	8.292	4.730	22.973	55.476	<0.001
attendance*T2	−92.522	<0.001	−1.008*1010	−92.522	−92.522	<0.001
0|1	0.886	0.606	1.464			0.143
1|2	1.896	0.540	3.512			<0.001
2|3	2.454	0.539	4.553			<0.001
3|4	2.982	0.557	5.358			<0.001

4. Discussion

This study bridges micro-evolutionary and macro-evolutionary approaches to identify the driving forces behind phenotypic diversity in avian nesting behaviour. By integrating comparative analyses across the bird phylogeny with behavioural experiments on the fitness consequences of nest wall variation, we reveal the mechanisms driving the correlated evolution of nest design, egg characteristics and body mass. Despite the strong correlation between body and egg mass, our findings indicate that egg mass, rather than body mass, acts as a primary selective pressure on nest wall construction. This supports the light-egg hypothesis that lighter eggs are more vulnerable to nest vibration and consequently require more substantial protection from nest walls. Our cross-species analysis confirms the evolutionary link between laying lighter eggs and the tendency to build nest walls, aligning with previous findings [9,15,16]. At micro-evolutionary level, our experiments demonstrate that nest walls directly enhance reproductive success by reducing the risk of egg falling off. By connecting the proximate nest function with ultimate evolutionary patterns, this study advances our understanding of bird nest evolution and establishes the critical role of nests in egg protection.

Nest walls represent a functional adaptation for retaining lighter eggs within nests. When comparing the absolute body mass, absolute egg mass and relative egg mass models, we found a consistent trend where decreasing absolute body and egg mass were correlated strongly with increasing likelihood of nest wall construction across bird phylogeny. This suggests that bird species producing lighter eggs, but not heavier-bodied species, experience stronger selective pressure to mitigate the risk of eggs rolling off. Additionally, our analysis of relative egg mass highlights how egg mass and nest site may interactively influence the evolution of nest wall construction. Specifically, non-ground nesters are more likely to build nest walls when they lay relatively lighter eggs, suggesting that these eggs in non-ground nests are less resistant to parent-induced nest vibration. Interestingly, while absolute egg mass exhibited a strong main effect on the likelihood of wall construction, relative egg mass lacked a comparable influence across all avian species. This may indicate that the egg-to-body mass relationship in birds has reached an evolutionary equilibrium within each type of nest site (ground/non-ground), where eggs of a given mass possess sufficient resistance to typical parent-induced nest vibration. Nest site, on the other hand, appears to influence nest wall construction only when considering the impact of nest vibration generated by parent birds (captured by relative egg mass). This suggests that the magnitude of nest vibration generated by parent birds can be modulated by nest site, and eggs in above-ground nests could face a higher risk of displacement due to increased instability.

Our experimental results align with our comparative analyses, supporting the light-egg hypothesis that lighter eggs are more prone to rolling off, making nest walls essential for their retention and improving egg-to-nestling survival. In nests without walls, we revealed higher rolling-off rates for lighter eggs (from lighter mothers), whereas eggs in nests with walls remained equally secure regardless of mass—demonstrating the adaptive value of this architectural feature. Our results reject the heavy-bird hypothesis, which predicts that heavier birds generate greater nest vibrations and consequently higher egg-rolling risk, driving nest wall evolution. Contrary to this prediction, the body mass of father birds had no significant effect on egg displacement risk, while heavier mother birds experienced lower egg rolling-off rates. Overall, these findings reinforce the idea that egg mass is the primary driver of nest wall evolution. They also provide empirical evidence that nest walls serve as a key evolutionary innovation, helping lighter birds persist in challenging nesting environments, such as unstable nest sites [14,20,47].

While egg mass is a strong driver of nest wall evolution, other traits, such as the amount of nest materials used and the duration of nest construction, were not associated with nest walls and are therefore less likely to influence or be influenced by their evolution. Our nest construction experiment of society finches reveals that nest walls, despite adding structural complexity, do not necessarily require more time or materials to build. This contrasts with previous comparative analyses suggesting that more complex nest structures require longer construction times [14]. This discrepancy may stem from either differences between inter- and intra-specific variation or substantial species-specific differences in nest-building investment [18]. If nest walls provide significant benefits for small birds without requiring greater time and energy investment, they may drive the evolution of more delicate nests. This could also promote behavioural complexity and cognitive enhancements for nest building [48]. These findings emphasize the importance of a behavioural perspective in explaining variations in nest and egg morphology [11].

Incorporating egg mass and nest attendance behaviour into bird phylogeny provides a valuable perspective on the interplay of multiple nest characteristics and the evolutionary trajectory of birds. During the rapid phylogenetic diversification, passerines have adopted various nest attachment strategies and nest locations—such as attaching nests laterally to reeds—that create previously unexploited nesting niches [20]. These strategies, however, may compromise nest stability [30], necessitating nest walls to secure eggs. Additionally, this evolutionary challenge is compounded by the altricial nature of passerines, whose smaller eggs require more intensive thermoregulation through increased parental nest attendance [49,50]. Although our experimental findings suggest that frequent nest attendance increases the rolling-off risk of eggs, it is also evident that nest walls play a crucial role in preventing egg loss, even under conditions of frequent attendance. Nest wall construction thus probably represents an effective evolutionary solution, enabling passerines to secure their lighter eggs within nests while maintaining necessary incubation behaviour. In addition, nest walls may improve thermoregulation by helping to maintain egg temperature during parental absences and enhance incubation efficiency when parents are actively incubating [51,52]. This thermoregulation function is particularly critical for small eggs, such as those of passerines, which cool faster than larger eggs [52]. Considering the roles of nest walls in safeguarding eggs and enhancing incubation efficiency enhances our understanding of the evolutionary relationships among various nest characteristics (e.g. nest structure, site and attachment) [20] and how these features collectively respond to multiple selective pressures acting simultaneously on avian reproductive strategies.

Our light-egg hypothesis, which proposes that nest walls offer functional advantages to birds with lighter eggs and higher nest attendance demands, complements two traditional hypotheses for enclosed nest construction: antipredation [28,29] and thermoregulation [51]. Together, these hypotheses provide a more comprehensive explanation for the latitudinal distribution of nest structure. Our hypothesis sheds light on the puzzle of why studies have failed to consistently observe a higher prevalence of enclosed nests in temperate regions for thermoregulation or in tropical zones for predation risk reduction [25]. For instance, we can now explain why natural selection may favour open nests in tropical passerines that face high predation pressures [29]. These birds may reduce nest predation by laying larger eggs, which require less nest attendance [28,49], thereby decreasing rolling-off risk and reducing demand for nest walls. Similarly, open nests may have been favoured, or at least not selected against, in temperate areas for species with larger eggs, as larger eggs excel at retaining metabolic heat [50] and tolerating lower incubation temperatures [49,53], thus negating the need for increased nest attendance or enclosed nests. Therefore, the light-egg hypothesis helps resolve the enigmatic latitudinal distribution of domed and cup nests.

Although we observed that nest walls reduced the rolling-off risk of eggs, it remains unclear whether higher nest walls provide proportionally greater protection. Future research should therefore shift the analytic focus from a binary variable (presence or absence of walls) to a continuous measure of wall height. However, quantitative data of this trait remain limited [13,17,54] and require more extensive investigations to quantify the relationship between wall height and egg security. Such studies would provide valuable insights into the optimal nest architecture for different species and ecological contexts.

Several factors beyond bird mass and egg mass may also influence rolling-off risk but were not directly examined in this study, including wind exposure and physical support of nest sites. Experimental investigations of wind exposure may further corroborate the light-egg hypothesis, as lighter eggs are more susceptible to movement in nests due to wind-induced vibration or displacement by strong wind, potentially observed in beach-nesting birds. Additionally, physical support structures that determine nest stability may influence rolling-off risk. For example, for nests built in trees, the thickness and number of supporting branches may be critical factors determining the degree of nest vibration and subsequent egg security. However, empirical evidence is required to confirm the effects of wind exposure and nest site support on egg retention.

Our experiments utilized cardboard boxes as nest sites and artificial nests, which differ from natural settings and may not fully reflect the selection forces operating in natural environments. Nevertheless, we believe that these designs are justified because all experimental nest sites share standardized spatial characteristics that control for confounding effects while isolating the physical forces of interest, such as gravity-caused nest vibration. Furthermore, the cardboard boxes and artificial nests in our experimental set-up exhibit vibration when birds land on them, mimicking the natural nest sites of tree branches or bushes used by the society finch’s ancestral species, the white-rumped munia (Lonchura striata) [55,56]. Future research could employ experimental settings more similar to natural conditions to test the light-egg hypothesis. For example, natural or artificial branches could serve as nest sites for experimental birds. However, such designs require careful standardization of branch size, structure and physical characteristics to control for experimental variation and confounding effects.

5. Conclusions

While our previous research suggests that more enclosed nests (e.g. nests with walls) may increase the risk of egg collisions, leading birds to invest more energy in producing eggs with more rigid shells [30], the functional advantage of egg containment probably outweighs this energetic cost. By linking macro-evolutionary trends with direct fitness consequences, this study identifies a main driver of nest wall evolution—egg mass. We provide cohesive evidence based on phylogenetic analyses and controlled experiments demonstrating that nest walls enhance egg containment, particularly in birds with lighter eggs. Without this integrated approach, disentangling the evolutionary mechanisms underlying the correlated evolution of body mass, egg mass and nest walls would be challenging. Birds actively engineer their reproductive micro-environments through nest construction, demonstrating a remarkable ability to manipulate environmental conditions to enhance offspring survival. The interaction among nest walls, vibration generated by nest attendance and nest site also highlights the importance of considering behavioural perspectives in understanding the evolution of nest morphology [47,57]. These findings illuminate how seemingly simple architectural innovations can dramatically expand species’ ecological opportunities, enabling successful colonization of otherwise inhospitable breeding environment and facilitating adaptive radiation in modern birds.

Ethics

All society finches used in this study were treated following a protocol approved by the Institutional Animal Care and Use Committee of Academia Sinica (protocol ID: 17-05-1096).

Data accessibility

We deposited the life-history character data of birds and codes used in this study to the Figshare repository [58].

Supplementary material is available online [59].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors’ contributions

C.-C.C.: conceptualization, data curation, formal analysis, investigation, methodology, visualization, writing—original draft; M.-N.T.: conceptualization, funding acquisition, methodology, supervision, validation, writing—review and editing; C.-M.H.: conceptualization, funding acquisition, methodology, project administration, resources, supervision, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This work was supported by Career Development Awards, Academia Sinica received by C.-M.H. and M.-N.T. (AS-CDA-108-L05 & AS-CDA-111-L07) and National Science and Technology Council, Taiwan received by C.-M.H. and C.-C.C. (NSTC 112-2311-B-001-039-MY3 & NSTC 113-2621-B-002-001-MY3).