Volume 90, Issue 5 p. 1122-1133
Free Access

Maternal allocation in relation to weather, predation and social factors in a colonial cooperative bird

Rita Fortuna

Corresponding Author

Rita Fortuna

CIBIO-InBIO - Research Centre in Biodiversity and Genetic Resources, Vairão, Portugal


Rita Fortuna

Email: [email protected]

Search for more papers by this author
Matthieu Paquet

Matthieu Paquet

Swedish University of Agricultural Sciences, Uppsala, Sweden

Search for more papers by this author
André C. Ferreira

André C. Ferreira

CIBIO-InBIO - Research Centre in Biodiversity and Genetic Resources, Vairão, Portugal

CEFE-UMR5175 CNRS - Université de Montpellier, Montpellier Cedex 5, France

Search for more papers by this author
Liliana R. Silva

Liliana R. Silva

CIBIO-InBIO - Research Centre in Biodiversity and Genetic Resources, Vairão, Portugal

Search for more papers by this author
Franck Theron

Franck Theron

CEFE-UMR5175 CNRS - Université de Montpellier, Montpellier Cedex 5, France

Search for more papers by this author
Claire Doutrelant

Claire Doutrelant

CEFE-UMR5175 CNRS - Université de Montpellier, Montpellier Cedex 5, France

Search for more papers by this author
Rita Covas

Rita Covas

CIBIO-InBIO - Research Centre in Biodiversity and Genetic Resources, Vairão, Portugal

Percy FitzPatrick Institute, DST-NRF Centre of Excellence, University of Cape Town, Cape Town, South Africa

Search for more papers by this author
First published: 07 February 2021
Citations: 7

Handling Editor: Albert Phillimore


  1. Females may adjust prenatal allocation in relation to ecological conditions that affect reproductive success, such as weather conditions or predation risk. In cooperative breeders, helpers might also influence reproductive success, and previous studies suggest that females can lay smaller eggs or larger clutches when breeding with more helpers. Although recent work suggests that helper effects can vary according to climatic variables, how social and ecological factors interact to shape prenatal allocation is poorly understood.
  2. Here, we examine how ecological and social components of the breeding environment covary with egg mass and clutch size, using as a model the sociable weaver Philetairus socius, a colonial, cooperatively breeding passerine. The study spanned 9 years and included over 1,900 eggs from over 550 clutches. Our analyses combined natural variation in weather conditions (rainfall before each reproductive event) with a nest predator-exclusion experiment and continuous monitoring of the mother's social environment, allowing us to estimate how individual females adjust allocation to reproduction as their number of helpers varies.
  3. We found that egg mass varied consistently within females and did not clearly differ in relation to rainfall or predation risk. Contrary to previous studies, there was no evidence for plastic adjustments as females gained and lost helpers, and egg mass was instead better predicted by mother size and identity.
  4. Females laid larger clutches when breeding in environments where predation risk was experimentally reduced and after higher rainfall levels. Yet, there was no evidence for increasing clutch size as the number of helpers increased, nor for an interaction between helper effects and ecological factors.
  5. We conclude that while sociable weaver females can vary their clutch size, they show high individual consistency in egg mass. In addition, we found no evidence that females may maximize fitness through plastic prenatal allocation in relation to the number of helpers, or that the presence/absence of helper effects is modulated by rainfall levels or predation risk. These results challenge our current knowledge on some of the possible benefits of breeding with helpers and call for more long-term analyses on reproductive allocation adjustments in other cooperative systems.


Life-history theory predicts that individuals should balance current reproductive investment against their future survival and breeding prospects (Stearns, 1992). These trade-offs are often modulated by a complex set of ecological conditions that define breeding environments (Erikstad et al., 1998; Stearns, 1992). Breeding females' investment strategies may involve adjusting prenatal allocation in response to ecological factors that have a predictable effect on reproductive success, thereby maximizing their reproductive output (Lindström, 1999; Mousseau & Fox, 1998).

Variation in climatic factors is among the set of ecological conditions that can influence reproductive allocation (Christians, 2002; Przybylo et al., 2000; Saino et al., 2004). In birds, favourable temperature and rainfall levels were found to associate with the production of larger eggs and clutches (Lepage & Lloyd, 2004; Schaper & Visser, 2013; but see Thomson & Hadfield, 2017). Another key ecological factor with a major influence on reproductive outcome is offspring predation risk (Martin, 1995; Zanette et al., 2011). Experimental manipulations of predation risk have revealed that mothers produced heavier eggs when nest predators were excluded (Fontaine & Martin, 2006) and laid smaller clutches when predator pressure seemed higher (Doligez & Clobert, 2003; Zanette et al., 2011). Saving energy when breeding in riskier environments, where progeny survival prospects are lower, can be advantageous over a lifetime if the immediate costs of reduced investment are counterbalanced by increased probability of breeding in the future (Mousseau & Fox, 1998; Stearns, 1992; Williams, 1966).

A species' social organization can also influence reproductive trade-offs, through variation in the number of conspecifics that are part of a colony or breeding group. Colonial living can benefit individuals by increasing foraging efficiency or reducing predation (Brown & Brown, 2001), but the resulting competition for resources can also be detrimental, affecting both maternal allocation and reproductive success (Bentz et al., 2013; Fuentes et al., 2019; Spottiswoode, 2007). In cooperatively breeding species, helpers assist breeding pairs by offering additional food to offspring, and larger groups are usually found to improve breeding environments (Cockburn et al., 2008; Downing et al., 2020). Hence, when breeding group sizes are predictable (Paquet et al., 2016), helpers' presence may trigger differential prenatal maternal allocation (Russell & Lummaa, 2009; Savage et al., 2015). Larger groups may allow females to raise additional offspring, which may explain the positive correlation between helpers and clutch size previously reported in, for example, apostlebirds Struthidea cinerea (Woxvold & Magrath, 2005) or red-winged fairy-wrens Malurus elegans (Lejeune et al., 2016; see also Liebl et al., 2016; Lloyd et al., 2009). However, this correlation is not found in all cooperatively breeding systems (Canestrari et al., 2011; Koenig et al., 2009; Russell et al., 2007; Santos & Macedo, 2011; Valencia et al., 2016), suggesting that the benefits of breeding with helpers may vary across species and/or environments.

Helper presence could also allow mothers to allocate differently to individual offspring, an idea that has been supported by findings of concealed helper effects on egg size (Dixit et al., 2017; Russell et al., 2007). Specifically, the load lightening hypothesis suggests that producing smaller eggs when breeding with helpers could improve mothers' survival without reducing offspring fitness, if this decrease in prenatal allocation is compensated by additional help to raise the young (Crick, 1992; Russell et al., 2007; Savage et al., 2015; Taborsky et al., 2007). This strategy was found in different taxa (Russell et al., 2007; Taborsky et al., 2007) and seems to be a general trend across cooperatively breeding species (Dixit et al., 2017). It should be favoured in species with higher future breeding probability, as individuals could maximize lifetime reproductive success through maximizing survival (Clutton-Brock, 1988). Alternatively, the differential allocation hypothesis (Sheldon, 2000) proposes that larger eggs are expected when breeding with more helpers, especially in cases where pre-birth care cannot be compensated by postnatal care or for species with a shorter reproductive life span (Russell & Lummaa, 2009; Savage et al., 2015).

However, most studies on the correlation between maternal allocation and helpers have compared differences among females (but see Lejeune et al., 2016), and therefore it remains unclear whether females plastically adjust their allocation to variation in their group size (Dixit et al., 2017). Alternatively, this correlation could be linked to habitat quality or to female traits, such as size, age or reproductive qualities, which are known to correlate with maternal allocation (Christians, 2002) and possibly also group size, if better quality females attract/produce more helpers (Cockburn, 1998; Dickinson & Hatchwell, 2004). Since manipulating helpers' number in wild systems is difficult, long-term studies provide a unique opportunity to study maternal allocation strategies. As the same female will likely gain and lose helpers over the years, it is possible to examine within-female plasticity in response to the number of helpers (Langmore et al., 2016; Lejeune et al., 2016). Moreover, the fitness benefits of adjusting allocation to helpers may depend on ecological factors that influence the success of current reproduction (Hatchwell, 1999). For example, a long-term study in superb fairy-wrens (Malurus cyaneus) found that females with helpers produced smaller eggs only under more favourable, cooler temperatures (Langmore et al., 2016). Conversely, in poor environments, the additive care of helpers and parents might be determinant for offspring survival (Hatchwell, 1999). It is therefore important to integrate multiple environmental variables and how they interact when studying reproductive strategies, which few studies have done (but see Koenig et al., 2009; Langmore et al., 2016; Lejeune et al., 2016).

Here, we examine how prenatal maternal allocation in a wild passerine relates to three key factors known to affect reproductive outcome—weather, predation risk and social environment. Our study spanned 9 years and combined a predator-exclusion experiment with natural variation in social and climatic factors. Repeated sampling allowed to disentangle fixed differences between females from allocation plasticity in relation to variation in number of helpers. We used as a study species the sociable weaver Philetairus socius, a colonial cooperative breeder inhabiting the arid Kalahari savannas, where variation in rainfall and temperature is extreme (Mares et al., 2017). These are relatively long-lived weavers (Paquet et al., 2015) that breed opportunistically, leading to long breeding seasons (Mares et al., 2017). However, reproductive success is low due to nest predation by snakes which can take 70% of all breeding attempts (Covas et al., 2008). For the surviving broods, a higher number of helpers correlates with higher feeding rates (A.C. Ferreira et al., unpubl. data; Covas et al., 2008), but breeding with helpers was only associated with improved fledgling condition during periods of low rainfall (Covas et al., 2008). Young female breeders, but not males, have shown increased survival associated with helpers' presence (Paquet et al., 2015). Importantly, in sociable weavers there is evidence that females assisted by helpers lay lighter eggs (Paquet et al., 2013), but whether this result is replicable across environmental conditions and represents plasticity has never been tested.

Given this species' life history, social behaviour and breeding environment, we made the following predictions for how maternal allocation strategies may vary with breeding conditions (Figure 1). In relation to ecological factors (weather and predation), we predicted that heavier or more eggs would be produced by mothers in favourable conditions, that is, higher rainfall/less extreme temperatures (Covas et al., 2008; Maclean, 1973b) and lower nest predation. However, in these favourable environments, mothers with more helpers are predicted to lay lighter eggs, in accordance with the load lightening hypothesis (Figure 1a), and to lay even more eggs than females without helpers (Figure 1b). Since we resampled females, we could test our hypotheses at the relevant biological level (within-female adjustments, see Dingemanse & Dochtermann, 2013), with the prediction that egg mass negatively correlated with the number of helpers within females, similar to between females (equal slopes; Figure 1c). Likewise, we expected a similar positive correlation between clutch size and number of helpers within and between females (equal slopes; Figure 1d).

Details are in the caption following the image
On the left, predicted interaction between number of helpers and breeding conditions on (a) egg mass and (b) clutch size. Solid lines represent good conditions (favourable weather/low predation) and dashed lines represent harsh conditions (adverse weather/high predation). For egg mass (a), we predicted lighter eggs as the number of helpers increases (load lightening), only under favourable ecological conditions. For clutch size (b), we expected larger clutches with more helpers, with a weaker effect in harsh environments. On the right, predicted direction of the correlation between number of helpers and egg mass (c) or clutch size (d) between females (solid line) and within females (dashed lines)


2.1 Study system

Sociable weavers are colonial cooperatively breeding birds endemic to southern Africa that build massive communal nests (Maclean, 1973a; colony size ranged from 3 to 134 individuals in this study). In our study population, most helpers (73%) are previous years' offspring of one or both breeders (A.C. Ferreira et al., unpubl. data; Covas et al., 2006), and helpers can assist one or several pairs with nestling feeding (Maclean, 1973c), nest building and sanitation (Ferreira, 2015; see Groups' identification for details on breeding groups).

This work was conducted at Benfontein Nature Reserve in Northern Cape Province, South Africa (28°520S, 24°500E), under permission from landowners, provincial authorities and the UCT Ethics committee.

2.2 Egg mass and clutch size

From 2008 to 2017, breeding monitoring was conducted in 20 different colonies (see protocol in Supporting Information section A). We weighed a total of 9,120 eggs and monitored 3,418 clutches. From these, we were able to sample 1928 eggs and 569 clutches for all the variables included in our analyses. Data estimates are reported for the latter dataset, which is also accessible in the Dryad Digital Repository archive. Sample sizes may vary between analyses of egg mass and clutch size due to missing parameters at specific breeding attempts (i.e. clutch size was known, but not all eggs were weighed).

2.3 Nest predation experiment

From 2010 to 2017, we conducted a nest predator-exclusion experiment. By wrapping tree trunks with heavy duty cling plastic film, we prevented snakes from climbing up the trees and reaching the colonies. In natural conditions, snakes forage conspicuously at the colonies, moving between nests (where they can spend several days), and sociable weavers might mob them, although rarely successful (pers. obs. from all authors). We protected eight colonies from snake predation for 1–6 years (Table S1). Four control colonies became protected and five protected colonies were posteriorly used as control. Among the resampled females (a total of 159; Table 1), 46 have experienced both treatments in different breeding seasons, 60 have only experienced natural conditions and 53 only bred in protected colonies (51 in natural and 49 in protected for the clutch size dataset).

TABLE 1. Number of female breeders partitioned by number of clutches sampled per female. The number of eggs sampled per female ranged between 2 and 22
No. clutches per female No. females
Egg mass dataset (N = 1,928) Clutch size dataset (N = 569)
1 94 100
2 60 54
3 42 41
4 27 30
5 15 9
6 14 11
7 1 1
No. resampled females 159 146
Total no. females 253 246

Our experiment decreased nest failure from 64% in control colonies to 44% in protected colonies (F = −12.5, p < 0.001; see Supporting Information section B and Table S2). Nest failure recorded after casual snake sightings was reduced from 35% in natural conditions to 13% in protected colonies. Fledgling success of each egg laid increased from 11% in control colonies to 25% in protected colonies (R. Fortuna, unpubl. data).

2.4 Groups' identification

We identified 107 breeding groups using direct observations and 507 from video recordings. When we had several video recordings per nest, mean number of helpers was estimated. Over the study period (2008–2017), 79% of the broods had at least one helper. Mean number of helpers per breeding attempt was 1.7 (SD = 1.3, range 0–6.7) and number of helpers was repeatable within the nestling period (R = 0.47; SE = 0.037; 95% CI = [0.36, 0.51]; p = 0.001; N = 1,523). To identify breeding pairs, we used a combination of genetic analyses (Paquet et al., 2015) and monitoring data. Our final dataset included 253 female breeders. The number of helpers' repeatability within resampled females was 0.21 (SE = 0.047; 95% CI = [0.126, 0.307]; p = 0.001; N = 520). Details on groups and breeders' identification are available in Supporting Information sections C–D.

2.5 Weather data

Weather data were provided by the South African Weather Service for the Kimberley Airport weather station (12 km from the study site).

Hot temperatures are known to affect birds' breeding success in this region (Cunningham et al., 2013) and sociable weaver's breeding seasons include the hottest months (December to February). In this study, most clutches (98%) were laid between September and April.

Rain is the major determinant of food abundance for this species (Maclean, 1973c), influencing their breeding onset and outcome (Altwegg et al., 2014; Covas et al., 2008; Mares et al., 2017). In this study, total annual rainfall ranged from 238 mm (2013) to 766 mm (2011).

We calculated three weather variables—total rainfall, mean maximum temperature and mean minimum temperature—over two short time windows of 30 and 15 days before laying dates, which were a priori chosen to represent short-term effects of weather variation experienced by females before laying each clutch. Periods shorter than 15 days were not included but were highly correlated with the used windows (Figure S1).

2.6 Statistical analyses

2.6.1 Climatic predictors

Due to limited knowledge of the climatic conditions that influence maternal allocation in this system, we tested which weather variable better improved the models by comparing Akaike information criterion (AIC) scores. We separately added each of the six weather variables to a model including all variables of interest and accounting for multiple other covariates likely to affect egg mass or clutch size (see below). We then selected the models with lowest AIC value for further inference. Models were fitted using maximum likelihood (ML) for comparison.

None of the weather variables clearly improved the egg mass model, although including total rainfall (mm) over 30 days before laying showed the lowest AIC model score (Table S3). Similarly, the best clutch size model included total rainfall (mm) over a 30-day period, but both rainfall and minimum temperature variables improved the clutch size baseline model (Table S3).

2.6.2 Effects on maternal allocation

To test whether helpers' effects on egg mass and clutch size were conditional on the predation experiment and/or weather conditions, we included two two-way interactions between these ecological variables and number of helpers. The weather variable used was total rainfall over the 30 days before laying (see above) and predation experiment was included as a binary factor (0–1 for natural conditions and protected colonies respectively). Remaining covariables aimed at controlling for allocation trade-offs and female attributes were as follows: clutch size or mean egg mass of the clutch (for egg mass and clutch size models respectively) and number of clutches previously laid by that female in that season (called ‘breeding attempt’; sociable weavers are multi-brooded), mothers' minimum age (in days) and tarsus length (Christians, 2002; Spottiswoode, 2007). Colony size was included as a proxy of breeding density, since egg mass was negatively correlated with this factor in previous studies (Spottiswoode, 2007). To account for repeated sampling of females and inter-seasonal and inter-colony variation, we included season, colony and female identity as random effects (in egg mass models, clutch identity was nested in female identity). All analyses were conducted using the R software v.3.6.0 (R Development Core Team, 2019).

Egg mass linear mixed models (LMM) assumed a normal distribution and were fitted by restricted maximum likelihood in lme4 (Bates et al., 2015; see Supporting Information section E). Clutch size models were under-dispersed using a Poisson error and we thus fitted both a LMM assuming a Normal distribution (see Section 3) and a cumulative link mixed model with clutch size as an ordinal categorical variable for comparison, both yielding similar results (see Supporting Information section E and Tables S6-S7).

For both response variables, numerical inputs were rescaled by subtracting the mean and diving by 2 SD (Gelman, 2008). Collinearity among predictors was assessed by calculating Spearman rank correlation coefficients (<0.52). Residuals' distribution and Normal distribution of random effects were assessed through diagnostic plots. Effects were considered significant when 95% confidence intervals did not overlap 0 and p values were lower than 0.05. Non-significant interactions were excluded to obtain final estimates, but no further model simplification was performed. Egg mass and clutch size repeatability within females was estimated by fitting the final models using the rptR package (Stoffel et al., 2017), which quantifies uncertainty in estimators by parametric bootstrapping (we used 1,001 bootstraps and 1,000 permutations; clutch size was log transformed for convergence). For each model, we present rescaled coefficients of numerical variables (see Table S4 for mean and SD). We also report marginal and conditional R2 (variance explained only by fixed effects and by both fixed and random effects respectively), calculated using the MuMIn package (Barton, 2009). Plots with raw data or predicted probabilities show untransformed numerical predictors.

2.6.3 Covariance partitioning

Univariate models allowed us to test if the general helper effect interacts with ecological conditions, but not to distinguish variation within females in relation to the number of helpers from fixed differences among females. To do this, we built mixed-effect models using egg mass and number of helpers, or clutch size and number of helpers, as bivariate responses. We used bivariate models instead of a subject-centring approach (Lejeune et al., 2016), to estimate female means as latent variables, and account for their uncertainty (credible intervals). This results in unbiased estimates of the within/between-individual effects, which could otherwise be substantially affected by measurement error (Lüdtke et al., 2008; see also Westneat et al., 2020). We performed a Bayesian analysis using the MCMCglmm package (Hadfield, 2010) with the same fixed effect structure and transformations as for the univariate models. The fixed effects were estimated on egg mass/clutch size, and an intercept was estimated for each response. A Normal distribution was used for egg mass, clutch size and number of helpers, the latter log transformed (0.5 was added before transformation to deal with zeros; Yamamura, 1999). Within and among females, we estimated a 2 × 2 matrix with a variance component for egg mass/clutch size, number of helpers and the covariance between each allocation variable and number of helpers, by fitting a random interaction between the bivariate response and each female observation or female identity respectively. We further added colony, season and clutch identities (for the model with egg mass) to the random structure. Estimates were obtained using vague priors (see Supporting Information section E for priors, model specification and convergence details). MCMCglmm was used to calculate posterior means with 95% credible intervals (highest posterior densities intervals or HPDs) for variances (V) and covariances (cov) estimated across thinned samples. We estimated a regression slope for the between-female (B) and within-female (w) random effects by dividing the estimated covariance by the number of helpers' variance for all posterior samples (Phillimore et al., 2010). The difference between the two slopes was as well-calculated (∆ slopes) from the posterior distributions. The 95% HPD of each slope and of their difference were used to determine whether slopes differed from zero and from each other, respectively, considering significant any credible intervals that did not include zero. Correlations between traits (r) were calculated by dividing traits' covariance by the square root of both traits' variance multiplied (Houslay & Wilson, 2017).


3.1 Egg mass

Egg mass ranged between 1.697 and 3.300 g (M = 2.518 g, SD = 0.195; N = 1,928; Table S4) with an adjusted repeatability within females of 0.502 (standard error SE = 0.034; 95% CI = [0.437, 0.568]; p = 0.001). Altogether, random effects explained most variation in egg mass, as a conditional R2 of 63% was obtained for the best model, but fixed effects alone explained only 4% of the variance (marginal R2) and showed quite small effect sizes on egg mass (Figure 2). Mother identity was the random factor explaining the largest variance (Mother ID variance = 0.019, Table S5).

Details are in the caption following the image
Standardized estimates and 95% CI of variables included in the egg mass linear mixed model. Variables of interest are placed first and remaining variables are ordered by effect size. Values indicate the effect on egg mass of a 2 SD change in numerical variables or from 0 to 1 in the predation experiment variable. Statistically significant effect is represented by filled circle

No clear effects of rainfall before laying were detected on egg mass (F = 1.72, df = 336.8, p = 0.09, estimate = 0.020 ± 0.012, 95% CI = [−0.003, 0.042]; Figure 2; Figure S3a; Table S5). The predator-exclusion experiment did not have a detectable influence on egg mass (F = 0.90, df = 142.3, p = 0.37, estimate = 0.013 ± 0.014, 95% CI = [−0.015, 0.042]; Figure 2; Figure S3c; Table S5), representing a minor mass increase of 0.5% in protected colonies compared to an egg laid in natural conditions.

We found no evidence for a correlation between egg mass and number of helpers (F = −1.30, df = 483.4, p = 0.19, estimate = −0.014 ± 0.011; Figure 2; Figure S3b; Table S5) and this was independent of the predation treatment or rainfall levels (Table S5; Figure S2a,b). There was also no evidence for a relationship between colony size and egg mass (F = −1.40, df = 30.8, p = 0.17, estimate = −0.027 ± 0.019; Figure 2; Table S5).

Egg mass variation seemed to depend on female body size as indicated by the tarsus length effect (F = 2.99, df = 235.7, p = 0.003, estimate = 0.059 ± 0.020; Figure 2; Table S5). There was no indication of an effect of clutch size on egg mass (F = −0.22, df = 463.5, p = 0.83, estimate = −0.002 ± 0.010; Figure 2; Table S5).

3.2 Clutch size

Over 90% of the clutches laid had two to four eggs (M = 3.2, SD = 0.6; N = 569; Table S4). Contrasting with egg mass results, mother identity did not clearly predict clutch size (variance mother ID = 0.019, df = 1, p = 0.12; Table S6) and there was no evidence for repeatability within females (R = 0.05; SE = 0.039; 95% CI = [0, 0.137]; p = 0.167). Clutch size was instead related to the random term ‘season’ (variance season = 0.044, df = 1, p = 0.002; Table S6) and there was a considerable proportion of variance unexplained by the model (residual variance = 0.287, R2 marginal = 0.074; R2 conditional = 0.246).

Higher rainfall levels were associated with larger clutches (F = 2.61, p = 0.01, estimate = 0.157 ± 0.06; Figure 3; Figure S5a; Table S6). Clutch size also differed with predation treatment, with a higher mean clutch size predicted in protected colonies (F = 2.31, p = 0.02, estimate = 0.137 ± 0.059; Figure 3; Figure S5b; Table S6).

Details are in the caption following the image
Standardized estimates and 95% CI of variables included in the clutch size linear mixed model. Variables of interest are placed first and remaining variables are ordered by effect size. Values indicate the effect on clutch size of a 2 SD change in numerical variables or from 0 to 1 in the predation experiment variable. Statistically significant effects are represented by filled circles

Clutch size did not clearly change as the number of helpers increased (F = −0.32, p = 0.75, estimate = −0.017 ± 0.053, Figure 3; Figure S5c; Table S6), and this was independent of rainfall levels (F = 1.20, p = 0.25; Table S6; Figure S4a) or the predation experiment (F = 1.39, p = 0.17; Table S6; Figure S4b). Colony size had no detectable effect on clutch size (F = −0.11, p = 0.91, estimate = 0–0.007 ± 0.062; Figure 3; Table S6).

Additionally, females laid larger clutches in latter reproductive attempts of the same season (F = 3.61, < 0.001, estimate = 0.211 ± 0.059; Figure 3; Table S6). There was no evidence that clutch size was correlated with mother size (F = 1.46, p = 0.15, estimate = 0.074 ± 0.051; Figure 3; Table S6) or with mean egg mass (F = −1.58, p = 0.12, estimate = −0.079 ± 0.05; Figure 3; Table S6).

3.3 Helper effects between and within females

The covariance analyses between allocation measures (egg mass and clutch size) and number of helpers, across and within females, revealed no credible evidence for helper effects (Table 2; Figure 4). In the bivariate model with egg mass (Table S9), we found a weak trend for a negative correlation between females (r = −0.078; CI = [−0.222, 0.077]) and an even weaker correlation within females (r = −0.039; CI = [−0.100, 0.029]), both with credible intervals overlapping zero (Table 2; Figure 4). There was no evidence that the between-female and within-female's slopes were credibly different from each other (∆ slopes = −0.013; CI = [−0.058, 0.036]; Table 2). Thus, we did not detect egg mass adjustments according to number of helpers, or fixed differences between females.

TABLE 2. Results from the variance–covariance matrices between and within females after modelling egg mass and number of helpers (top), and clutch size and number of helpers (bottom), as bivariate responses. Mean estimated variances (V) are presented on the diagonals, correlations (r) above and covariances (cov) below. The posterior mean of regression slopes between (B) and within females (w), and their difference (∆ slopes), are also presented. After each value, 95% credible intervals are shown
Egg mass No. helpers Slope (cov/Vh) ∆ slopes (B-w)
Between females (mother ID) Egg mass V e  = 0.03 (0.024, 0.036) r = −0.078 (−0.222, 0.077) −0.023 (−0.07, 0.02) −0.013 (−0.058, 0.036)
No. helpers cov = −0.008 (−0.022, 0.009) Vh = 0.339 (0.274, 0.411)
Within females (residuals) Egg mass Ve = 0.016 (0.015, 0.017) r = −0.04 (−0.1, 0.0.29) −0.01 (−0.025, 0.007)
No. helpers cov = −0.0026 (−0.006, 0.002) Vh = 0.263 (0.245, 0.28)
Clutch size No. helpers Slope (cov/Vh) ∆ slopes (B-w)
Between females (mother ID) Clutch size Vcs = 0.077 (0.05, 0.104) r = 0.040 (−0.244, 0.281) 0.027 (−0.167, 0.197) 0.043 (−0.184, 0.24)
No. helpers cov = 0.005 (−0.027, 0.035) Vh = 0.171 (0.108, 0.233)
Within females (residuals) Clutch size Vcs = 0.271 (0.238, 0.309) r = −0.119 (−0.113, 0.092) −0.016 (−0.098, 0.074)
No. helpers cov = −0.006 (−0.038, 0.029) Vh = 0.393 (0.336, 0.449)
Details are in the caption following the image
Correlation between number of helpers and egg mass (left side) and clutch size (right side), between females (black full line bars) and within females (grey-dashed line bars), calculated from respective bivariate models (Table 2). Circles show posterior mean correlations and vertical bars represent 95% credible intervals. Within- and between-female slopes were not credibly different (see Table 2)

For the relationship between clutch size and number of helpers (Table S10), our estimates showed opposite but unclear correlations when comparing between- and within-female trends (Table 2). The posterior mean correlation of clutch size and number of helpers between females was slightly above zero (r = 0.039; CI = [−0.243, 0.281]) and within females was below zero (r = −0.02; CI = [−0.113, 0.092]), both with credible intervals overlapping zero (Figure 4). Similarly, there was no indication that slopes between and within females were credibly different from each other (∆ slopes = 0.043; CI = [−0.184, 0.240]; Table 2).


In this study, we investigated how maternal allocation relates to females' breeding environment in a cooperatively breeding species. Our results show that egg mass did not clearly correlate with weather conditions, experimentally reduced nest predation or the size of social groups. Clutch size was flexible within females and positively associated with higher rainfall levels, and experimentally reduced nest predation. Females did not show egg or clutch size adjustments when breeding with more helpers and ecological conditions at laying were not found to modulate helper effects on maternal allocation. We thus found no evidence for fixed or plastic prenatal reproductive strategies in relation to number of helpers in sociable weavers.

4.1 Number of helpers and within-female allocation

Contrary to expected, we found no indication that sociable weaver mothers adjust prenatal allocation to the number of helpers in their group. Similarly, a previous long-term study that investigated plastic responses to number of helpers did not find egg size adjustments, but females increased clutch sizes when breeding in larger groups (Lejeune et al., 2016). In our study, covariance partitioning analyses showed that mothers do not seem to benefit from the presence of helpers by plastically load lightening or producing more offspring when breeding in larger groups. These egg mass results differ from previous work in this population (Paquet et al., 2013) and challenge the overall evidence for prenatal load lightening in cooperatively breeding systems reviewed in Dixit et al., (2017). In fact, this effect was mainly driven by three species (Canestrari et al., 2011; Paquet et al., 2013; Taborsky et al., 2007 one fish species) and after updating Dixit et al.'s (2017) analysis with the effect obtained here, we found no general tendency of load lightening at the egg stage in cooperative breeders (estimate = −0.1320, CI = [−0.2921, 0.0281]; p = 0.1061; Figure S7; details in Supporting Information section F). Moreover, three additional studies have recently reported no evidence of prenatal load lightening in other cooperative breeders (Cusick et al., 2018; Van de Loock, 2019; Zhao et al., 2019). The meta-analysis on prenatal load lightening in cooperative breeders should therefore be revisited in a future investigation. That the results obtained here contrast with a previous 1-year study in this species (Paquet et al., 2013) demonstrates the importance of replicating short-term investigations, as these might provide limited insights of evolutionary processes acting on natural populations in fluctuating environments (Cockburn, 2014; Fargevieille et al., 2017; Langmore et al., 2016). Furthermore, our work shows that clutch size does not clearly correlate with helpers' number in sociable weavers, concurring with investigations on several other species (Canestrari et al., 2011; Koenig et al., 2009; Russell et al., 2007; Santos & Macedo, 2011; Valencia et al., 2016). The failure of social factors to predict maternal allocation extended to colony size, our proxy of breeding density, which did not clearly correlate with egg mass or clutch size (but see Spottiswoode, 2007).

Our analyses relied on natural variation in group sizes and, if variation sampled within females is small, helper effects could be harder to detect. However, we found low repeatability in number of helpers for individual females (R = 0.2), suggesting that little within-female variation is an improbable cause for failing to detect helper effects. Lack of plasticity in maternal allocation could also arise if females cannot predict the amount of help they expect to receive (Russell et al., 2007), but previous work in our population suggests that mothers have reliable cues regarding their number of helpers, since most helpers are previous offspring of the breeders (73%; A.C. Ferreira et al., unpubl. data; Covas et al., 2006), social bonds are stronger within breeding groups (Ferreira et al., 2020) and roosting groups before breeding were correlated with breeding group sizes (Paquet et al., 2016). However, in sociable weavers, there appears to be substantial within-individual variation in the amount of help provided (A.C. Ferreira et al., unpubl. data) and further assessments of helping behaviour repeatability within and across broods would help to understand which cues are available for mothers prior to laying.

Long-term investigations have suggested that helper effects on female allocation may be detectable only under favourable climatic conditions (Langmore et al., 2016) and we therefore examined the effects of weather, namely rainfall, in interaction with number of helpers. This was especially relevant given that evidence for prenatal load lightening in sociable weavers had been found during the season with the highest total annual rainfall in our dataset (Paquet et al., 2013). Yet, our findings suggest that the likelihood of observing helper effects on maternal allocation does not seem to be determined by climatic conditions, unlike what was shown in superb fairy-wrens M. cyaneus (Langmore et al., 2016).

4.2 Response to nest predator exclusion

By decreasing actual nest predation rates by snakes, we increased brood survival in manipulated colonies, which together with the decrease in snake foraging activity (indirect predation effects) was expected to increase mothers' allocation to reproduction. Our results indicate that females responded to these cues as they laid larger clutches in protected colonies than in natural conditions. This concurs with previous studies that reported adjustments in clutch size in response to offspring predation risk (Doligez & Clobert, 2003; Julliard et al., 1997; Zanette et al., 2011). In contrast, and unlike a previous study (Fontaine & Martin, 2006), egg mass was not clearly affected by our predator-exclusion experiment. For both egg mass and clutch size, helper effects did not detectably differ across predation treatments.

Larger clutches in protected colonies could have been caused not by reduced predation risk per se, but by females being in better condition. Mothers in protected colonies could save energy by laying less clutches when compared to females in natural conditions (which suffer higher predation and hence lay more replacement clutches). However, our variable ‘breeding attempt’, which accounts for the number of clutches previously laid that season, showed no evidence of a negative effect on females' fecundity. Females' response to reduced predation could also arise from assessing self or conspecific breeding success instead of predation risk (reviewed in Ibáñez-álamo et al., 2015). Here, we cannot determine whether the mechanism underlying females' response is a decrease in perceived or actual predation risk. Nevertheless, our results suggest that responding to safety cues can be as much an adaptive mechanism as responding to danger cues (i.e. increased predation; Luttbeg et al., 2020). Sociable weaver females therefore seem to assess the quality of their breeding environment and increase the number of offspring produced when the expected value of the current breeding attempt is higher (Mousseau & Fox, 1998; Stearns, 1992).

4.3 Weather effects

Females laid larger clutches after higher rainfall levels, which represents favourable conditions in these arid habitats (e.g. Covas et al., 2008; Maclean, 1973b; Mares et al., 2017; see also Aranzamendi et al., 2019; Lloyd, 1999). Moreover, there was no evidence for within-female clutch size repeatability and mothers showed a clear tendency for producing larger clutches in latter breeding attempts of the breeding season. These climatic and seasonal correlations are likely due to an increase in resources following summer rainfall peaks (Dean & Milton, 2001), which is expected to improve females' condition and allow them to raise more young.

Unlike clutch size, egg mass did not clearly correlate with rainfall levels before laying. This may indicate that egg mass is not highly dependent on resource availability in this system, a result that has also been reported in other birds (Christians, 2002; Thomson & Hadfield, 2017).

4.4 Egg mass consistency

The wide range of egg mass values was mostly predicted by mother identity and body size, which concurs with previous results on low levels of egg size variation across ecological conditions and high consistency within females (Christians, 2002; Griffith et al., 2020). The effect of ‘mother identity’ may be explained by genetic features that define the amount of resources that each female allocates to her eggs (Christians & Williams, 2001). Future estimates of heritability and fitness differences between females could help explaining the adaptive causes of egg mass consistency in our system (Christians, 2002).

Mother identity effects could also be influenced by the identity of their breeding partner. Sociable weavers exhibit long-term monogamy (P.B. D'Amelio et al., unpubl. data) and females' prenatal allocation can vary with male quality proxies (Horváthová et al., 2012). Additionally, we found substantial egg mass residual variance, which might be attributed to intra-clutch differences according to the laying sequence (Kozlowski & Ricklefs, 2010).


We found that sociable weavers' clutch size varies with rainfall and predation, two ecological factors known to impact different aspects of this species' breeding biology, and suggesting an opportunistic strategy to maximize reproductive output in their highly variable ecosystem. In contrast, egg mass was consistent within females and across ecological conditions. Unexpectedly, we found no evidence for clutch size or egg mass plasticity in relation to number of helpers. These results challenge our current understanding of helper effects in our system and other cooperatively breeding species. The present results, together with the large variation in egg mass generally found among females (Christians, 2002; this study), highlight the value of testing within-individual differences, as well as the importance of repeated sampling across variable environments when studying reproductive strategies.


We are grateful to the field assistants and volunteers who collected field data over the years and to Pietro D'Amelio and Céline Teplitsky for useful discussions on data analysis. We thank the editors and three anonymous reviewers for constructive comments on previous versions of this manuscript. De Beers Consolidated Mines gave the permission to work at Benfontein Reserve. The South African Weather Service (SAWS) provided weather data. We thank FitzPatrick Institute of African Ornithology (DST-NRF Centre of Excellence) at University of Cape Town (South Africa), FCT (Portugal) for funding through grants IF/01411/2014/CP1256/CT0007 and PTDC/BIA-EVF/5249/2014 to R.C., French ANR (Projects ANR 15-CE32-0012-02 and ANR 19-CE02-0014-02) to C.D., OSU OREME and Marie Curie-IRSES 318994. We thank Ben Hatchwell for part of the molecular sexing and genotyping, conducted at the University of Sheffield under grant NE/K015257/1. This work was conducted under the CNRS-CIBIO Laboratoire International Associé (LIA). R.F. and A.C.F. were funded by FCT (SFRH/BD/130134/2017 and SFRH/BD/122106/2016).


    R.C. and C.D. conceived the study; R.F., A.C.F., L.R.S. and F.T. extracted and compiled data; R.F. analysed the data with input from M.P., R.C., C.D. and A.C.F.; R.F. led the writing of the manuscript with input from M.P., C.D. and R.C.; All the authors collected the field data, gave feedback on the manuscript and their final approval for publication.


    Data available from the Dryad Digital Repository https://doi.org/10.5061/dryad.jm63xsj97 (Fortuna et al., 2021).

      Journal list menu