Early differences in maternal food allocation predict later aggression toward dependent offspring in domestic canaries
Description
Maternal aggression (MA) toward offspring is typically observed near nutritional independence and interpreted as conflict over care duration. In domestic canaries, however, MA arises earlier, when offspring are still dependent on parental provisioning, raising the question whether it reflects differences in maternal allocation strategy or response to offspring-driven interactions. We examined whether females that later express MA already differ in how they allocate food before aggression begins. Midway through the nestling period, we experimentally manipulated hunger to test maternal responsiveness to short-term need, nestling body mass and begging behaviour. MA and noMA females did not differ in initial reproductive investment or overall provisioning rates. However, they differed in how they weighted offspring cues: although all females responded to hunger, MA females showed a stronger association between provisioning and nestling body mass and integrated behavioural cues differently, resulting in more biased allocation among siblings. Within MA nests, nestlings later exposed to aggression already received fewer food transfers, yet did not differ in body mass or begging behaviour prior to aggression. In addition, MA females lost less body mass across the breeding period. Together, these findings indicate that MA is preceded by structured differences in early maternal allocation and cost management rather than differences in overall effort or reactive responses to offspring demand.
Files
Steps to reproduce
ll analyses were conducted in R v4.3.0. For each research question, we specified a biologically informed full model including all relevant fixed effects and interactions, and simplified models by backward elimination of non-significant interactions. Linear models (LMs) and linear mixed models (LMMs; lme4 v1.1-35.2) were used for continuous behavioural responses, and generalised linear models (GLMs) and generalised linear mixed models (GLMMs; glmmTMB v1.1.7) for provisioning data. Provisioning models were fitted with a negative binomial error distribution (nbinom1 or nbinom2, selected based on dispersion diagnostics and model fit). Nest identity was included as a random effect where repeated measures per brood occurred. Model assumptions and dispersion were assessed using DHARMa (v0.4.6). Significance was set at α = 0.05, and p-values for GL(M)Ms were obtained from Wald tests.
Institutions
- University of AntwerpFlanders, Antwerp