The resource-tracking/facultative brood reduction hypothesis suggests that, under food stress, many altricial birds sacrifice the youngest brood members to enhance the growth and survival of their siblings. Studies examining staggered hatching and food shortage have generally supported this idea, although staggered hatching may serve additional purposes. However, the direct beneficiaries of this selective mortality remain unclear, as most research has overlooked parents and post-fledging outcomes. A life history perspective has rarely been applied to brood reduction. Using a 34-year individual-based dataset on blue-footed booby (Sula nebouxii) reproduction, we investigated whether siblicidal reduction of food-stressed two-chick broods benefits mothers, fathers, or surviving offspring. Results revealed that mothers of reduced broods were ~16% more likely to survive to the next breeding season than mothers of intact broods, indicating a significant maternal benefit. Moreover, cessation of sibling competition allowed surviving chicks to achieve fledging body condition, breeding probability, longevity, and lifetime reproductive success comparable to chicks from intact broods. Thus, sacrificing the subordinate chick ensures adequate—not privileged—growth of the surviving sibling and enhances maternal survival, while apparently providing no survival benefit to fathers. However, experimental confirmation is required to validate these findings and further explore the underlying mechanisms.

Colony monitoring

 Between 1989 and 2023, the contents of all booby nests in two study areas of Isla Isabel, off the Pacific coast of Mexico (21°50’59”N, 105°52’54”W), were monitored every 3 days throughout the first three months of every five-month breeding season and every 6 days subsequently (details in Drummond et al., 1986, 2003). Hatchlings were labelled according to hatching order with coloured leg-wires, then numbered plastic bands, and at 70 days (a proxy of fledging, which can occur as early as 80 days) they were weighed, measured (culmen and ulna), and fitted with alphanumeric steel bands. Sex of breeders was identified by voice (females grunt, males whistle) but sex of chicks was normally unknown until they recruited into the breeding population a few years after fledging. Due to limited natal dispersal in this population (30.5 ± 1.7 m in males and 36.6 ± 1.4 m in females; 𝑥 ± 𝑠𝑒) and lifetime philopatry (Kim et al. 2007), this monitoring captured the development and fledging of every clutch and brood, along with the recruitment, annual survival and reproductive performance of offspring of all but the most recent cohorts across their life spans.

Benefits to parents

We evaluated whether parents that experienced brood reduction (one chick died) outperformed parents of the same sex that successfully fledged both chicks, in terms of survival, early laying, and breeding success in the subsequent year. Our sample was all reproductive events between 1993 and 2022 involving known-age parents (banded at fledging) with a brood of two hatched chicks. First, we compared the probability of surviving to the subsequent breeding season between parents that experienced brood reduction versus those that fledged both chicks. Death of a parent was inferred when an individual failed to breed in the study areas during five consecutive years, and ascribed to the last year it bred (only 2% of breeders take more than five gap years; Drummond et al., unpublished data). Second, for parents that survived, we compared laying dates and breeding success, measured as the number of fledglings produced, in the subsequent year between those that experienced brood reduction and those whose broods fledged intact.

Benefits to chicks

Using all two-chick broods that hatched between 1989 and 2023, we first compared the fledging body condition (next paragraph) of fledglings from two-chick broods hatched between 1995 and 2023, contrasting individuals of the same hatch order from reduced versus intact broods. We also compared fledglings from reduced versus intact two-chick broods that fledged between 1995 and 2011 for recruitment success (whether they cared for a clutch within the first 12 years of life), longevity, and lifetime reproductive success (total fledglings produced). Recruitment after 12 years is rare (Drummond et al., 2011). In addition, for males and females that recruited and died between 1999 and 2017, we compared age at recruitment, longevity, and lifetime reproductive success of individuals of the same sex from intact versus reduced broods.

Body condition at fledging was assessed using a scaled mass index, which adjusts body mass relative to body size, thereby accommodating fluctuations in the body mass-size relationship of each individual over its life span (equation 1; Peig and Green 2009). Senior fledglings from intact and reduced broods weighed 1665 ± 238 grams and 1549 ± 243 grams, respectively, with ulna lengths of 205 ± 11 mm and 205 ± 11 mm[HD1] , respectively. Junior fledglings from intact and reduced broods weighed 1614 ± 245 grams and 1560 ± 236 grams, respectively, with ulnas measuring 204 ±11 mm and 203 ± 11 mm, respectively.

Equation 1. Scaled mass index

 denotes the scaled mass index value,  is body mass (grams), L₀ is the population’s mean ulna length (mm),  is the individual ulna length, and b is the scaling exponent, estimated as the slope of the Standardized Major Axis (SMA) regression of the log-transformed body mass on ulna length.

Brood reduction, whether by starvation or attacks by adult neighbours on chicks seeking adoption, respectively, was inferred when monitors found a chick older than 7 days dead on the parents’ territory, or absent from that territory and its surroundings. When both chicks died within nine days of each other (within 3 consecutive nest checks), their deaths were attributed to nest abandonment. Disappearance of chicks younger than 7 days from the nest was attributed to predation by milk snakes (Lampropeltis polyzona), the only local predator of chicks (Ortega et al. 2021). Only chicks younger than 7 days are small enough to be consumed by snakes, given the constraints of their mandibular gape. While sibling aggression begins around 5–6 days old (Drummond et al., 1991), brood reduction by expulsion is unlikely at this stage since younger chicks lack the locomotor ability to leave the nest. If, however, siblicide by the older sibling occurred, the chick’s body would have been found in the nest during monitoring and categorized as brood reduction. Lifetime reproductive success was the number of fledglings produced by each male or female breeder over its complete breeding history.

Note that sample sizes in Results differ between parent and chick analyses wherever (1) chicks were excluded because missing morphological or body mass data prevented calculation of their scaled mass index, or (2) parental identities were unknown.

Statistical analyses

Benefit to parents.

Models evaluating the benefits of brood reduction to parental mortality (i.e., the probability of surviving to the next breeding season) and reproductive performance (i.e., number of fledglings produced) included a two-way interaction between brood reduction (yes, no) and sex (male, female). Both the linear and quadratic expressions of the parent’s age were included as covariables to accommodate quadratic age-related changes in the boobies’ performance (Torres and Velando 2007; Beamonte-Barrientos et al. 2010). Parental identity, nest number, and current year were included as random effects; the first two of these factors accounted for statistical non-independence, while the year accounted for unmeasured environmental variables during the breeding season (e.g., food availability, parental provisioning, population density; Ancona and Drummond 2013). To measure the influence of brood reduction on the probability of surviving to the subsequent year, we constructed Bernoulli generalized linear mixed models (GLMMs) with logit link function. To analyse laying date and breeding success (number of fledglings produced), we used a LMM with an identity link function and a Poisson GLMM with a log link function.

Benefit to chicks.

Our models assessed differences in life history and reproductive outputs between individuals from intact broods and those from reduced broods. All models incorporated brood reduction and hatching order (senior, junior) as fixed effects, and birth cohort and nest number as random effects. Birth cohort accommodated unmeasured environmental factors during the recruits' natal years which could affect sibling competition (Ancona and Drummond 2013), while nest number controlled for statistical non-independence when two siblings were included in the same model.

To compare the influence of brood reduction on body condition at fledging and recruitment probability we built two GLMMs.  For body condition we employed a Gaussian model with an identity link function, and for recruitment probability we fitted Bernoulli models with a logit link function. In these models we examined a three-way interaction involving brood reduction, hatching order and hatching date (standardized with November 3rd as day 1), the latter to account for intraseasonal weather variations. Inclusion of hatching date was crucial because as the breeding season progresses on Isla Isabel, rainfall decreases, sea surface temperature rises, and primary productivity of the ocean declines, all of which collectively contribute to increased mortality of chicks hatched late in the season (Ortega et al. 2022). The three-way interaction explored whether the possible benefit of brood reduction to the surviving sibling increases with lateness of hatching, as adverse environmental conditions later in the season potentially intensify the antagonistic interactions between competing siblings.

We compared effects of brood reduction on age at first reproduction, longevity and lifetime reproductive success (total number of fledglings produced) of recruits (all of which were sexed) using Poisson models with a log link function. These models examined a three-way interaction involving reduction of the natal brood, sex, and body condition at fledging to test for possible sex-related and long-term effects of nestling growth following brood reduction. The longevity and lifetime reproductive success models included age at first reproduction as a fixed effect. Additionally, lifetime reproductive success was included as a fixed effect on the longevity analysis, while longevity was added as an independent variable to the lifetime reproductive success models. Age at first reproduction was included because late recruitment could reduce the number of breeding years available to a booby; longevity controlled for selective disappearance (van De Pol and Verhulst 2006); and total number of fledglings produced controlled for possible effects of the trade-off between reproduction and somatic maintenance.

All analyses used the R statistical environment (R Development Core Team. 2024). Prior to model fitting, continuous fixed variables were z-score normalized to enhance the interpretability of parameter estimates (Grueber et al. 2011; Cade 2015). To ensure computational stability, prevent overfitting, and mitigate the risk of erroneous estimations of large effect sizes, weakly informative priors were incorporated into all models (Gelman et al. 2008; Lemoine et al. 2016). For all fixed effects in the models, a normal prior of N(0,1) was assigned, implying an expectation that the majority of responses would fall within one standard deviation of the average response values, making large effects an unusual occurrence (Lemoine et al. 2016). Variable standardization was performed using the built-in scale function (R Development Core Team 2023), and the brms function within the brms package was used to fit all models (Bürkner 2023; Stan Modeling Language Users Guide and Reference Manual, 2.35 2024). For each dependent variable, we calculated the contrast of interest between the estimated mean effects using the emmeans package (Lenth et al. 2022).

The posterior distributions of the parameters, accompanied by their 89% highest posterior density intervals (HPD), representing high-probability interval of parameter values (McElreath, 2020), were drawn by executing five randomly initiated Markov chains. Each chain underwent 10,000 iterations, with a burn-in period of 1000 iterations. Posterior predictive checks were performed for each generalized linear mixed models (GLMMs) using the launch-shinystan function of the shinystan package (Gabry Jonah et al. 2022). Complete model summaries are provided in Supplementary Material. Models that exhibited divergent transitions were excluded to avoid unreliable inferences.