Balancing care and conflict: towards a better understanding of maternal aggression in canaries (Dataset)

Description

Parental care improves offspring quality and survival and hence enhances parental fitness, still seemingly maladaptive parental behaviours such as directed aggression and neglect have been reported for a variety of species. While maternal aggression may represent a behavioural pathology, it could also be an adaptive strategy enabling optimal resource allocation in the context of a trade-off between current and future reproduction and other evolutionary conflicts of interest. This study evaluated the functional consequences of maternal aggression for developing offspring in the domestic canary (Serinus canaria). We found that maternal aggression significantly impaired offspring growth, but it did not reduce their survival, thus it might have lasting negative effects on offspring quality. Additionally, early-life exposure to aggression resulted in increased threatening behaviours in juvenile male but not female offspring, which again suggests that understanding the significance of an aggressive parenting style may require incorporating long-term effects. Females exhibiting maternal aggression did not lay larger second clutches, as would have been expected if aggression relates to a trade-off between current and future reproduction. Still, they laid overall larger clutches that were less variable in size. This indicates that there might be phenotypic differences, with females behaving aggressively towards their offspring being less flexible and too optimistic at egg laying. Thus, despite having negative effects on offspring growth parental aggression could nevertheless represent an adaptive, context-dependent behavioural strategy.

Steps to reproduce

All statistical analysis were performed in R 4.1.3. Backward elimination for non-significant interactions (α = .05) was used to build the minimal models using lmerTest (v3.1.3) to calculate p-values (by default this package provides a type-III Anova). The significance level was set at α = .05. To inspect for violations of the model assumptions, lmm plots were generated using the packages performance (v0.10.3) and DHARMa (v0.4.6). To test the effects of maternal aggression (MA) on growth we fitted a linear mixed effects model (lme4 package, v1.1.33) with body mass as the response variable and age (included as a polynomial 4th order) as an explanatory variable as this was the simplest model to explain the relationship between weight and age. We also included MA, hatching order, year, and nestling sex, as well as the interaction between maternal aggression and age, and the interaction between nestling sex and age. Random intercepts were added for nest id and chick id nested within nest id, as well as a random slope term for age (again, included as a 4th order polynomial). To test the effect of MA on offspring survival we fitted a Cox proportional hazards (survival package; v3.5.5) model including MA, hatching order, body mass at hatch, and year as predictor variables. To test the effects of MA on the offspring behaviour, we fitted a linear mixed-effects model for the time spent threatening including MA, nestling sex, and the interaction between both terms. Time threatening, was square-root transformed. Nest id was added as a random-effect. Direct aggression was analysed with a Fisher exact test. We used maternal aggression (yes/no), and direct aggression (0/1). To test the relationship between MA and parental investment, clutch size was analysed in a linear mixed-effects model as the count data followed a normal distribution. In this model, we included as fixed effects MA, year, clutch, and the interactions between MA and year and MA and clutch. Nest id was included as a random effect. A Levene’s test (car package, v3.1.2) was used to compare the overall variation in clutch size between females that exerted MA and those who didn’t. Second clutch mass was analysed using a linear model with maternal aggression, brood size of the first brood at day 10, year and the pairwise interactions as fixed effects on the full model. This model was run using robust standard errors (type HC1) via the coeftest function of the lmtest package (v0.9.40). The time interval between laying the first and the second clutches was analysed using a linear model with the same set of predictors and interactions used for the second clutch mass analysis

Institutions

Categories

Funding

Licence