No evidence for a link between maternal aggression and inhibitory control across generations in domestic canaries (Dataset)
Description
This dataset contains behavioural data collected from domestic canaries (Serinus canaria) tested in a detour task designed to assess inhibitory control (response inhibition). Data were collected at the University of Antwerp (Belgium) during the 2024 breeding season as part of a larger transgenerational study on maternal aggression (MA) in this species. The dataset includes detour task performance records for two groups: adult breeding females tested before the breeding season, and their juvenile offspring tested at approximately 44–45 days of age. A cross-fostering design was applied at clutch completion, allowing biological and foster mother effects on offspring performance to be statistically disentangled. For each individual and trial, the dataset includes detour latency (seconds from test box entry to barrier circumvention), total time spent persisting at the barrier (cumulative seconds inside the barrier zone of interest), and a multi-baseline motivational measure (mean latency to reach the food bowl across habituation trials 2 and 3 with an opaque barrier). Individual identifiers, nest of origin, foster nest, maternal aggression phenotype of the biological and foster mother (binary: MA / noMA), sex, age class (adult / juvenile), test session (1 or 2), trial number (1–3 within session), and barrier type (horizontal / vertical) and order are also included. Maternal aggression was defined as feather plucking directed towards at least one nestling, scored from video recordings of nest checks conducted daily from day 14 post-hatch. Baseline food-approach motivation for adult females (latency to feed during the first habituation trial) is provided as a supplementary variable. All raw values are included without winsorisation or transformation; any processing steps applied during analysis are described in the reproducibility notes below.
Files
Steps to reproduce
All statistical analyses were conducted in R (version 4.1.3; R Core Team, 2013). The following steps describe how to reproduce the main results reported in the associated manuscript (Garcia-Co et al., in prep., University of Antwerp). 1. Software and packages. Install the following R packages: lme4 (v1.1-35.5), lmerTest (v3.1.3), glmmTMB (v1.1.12), car, DHARMa (v0.4.6), and performance (v0.10.3). 2. Data preparation. Load the dataset and apply exclusion criteria: remove individuals that did not approach the food bowl by the third habituation day, those that neither detoured nor entered the barrier zone of interest during testing, and those that left the test box without contacting the food bowl after detouring. Winsorise detour latency values using a trim value of 0.05 (per individual across trials) prior to modelling. Log-transform winsorised latency values. For adult persistence, compute the proportion of trial time spent persisting; for offspring, use absolute seconds persisting. Adjust proportion values of exactly 0 or 1 by the standard beta regression correction ((y · (n−1) + 0.5) / n) prior to beta modelling. 3. Prediction 1 (adult females). Fit a linear mixed-effects model (LMM) with log-transformed winsorised detour latency as the response, MA phenotype and multi-baseline measure as fixed effects, Trial (1–6, continuous) as a covariate, and female identity as a random intercept. Fit a beta mixed-effects model (glmmTMB, logit link) with the same fixed and random effects structure for proportion of time persisting. 4. Predictions 2a and 2b (offspring). Fit LMMs with biological mother MA phenotype, foster mother MA phenotype, their interaction, multi-baseline measure, and Trial as fixed effects; individual identity nested within nest as random intercept. For total time persisting, use an LMM with the same structure. 5. Age class comparison. Fit LMMs including age class (adult / juvenile), Trial, and their interaction as fixed effects; individual identity nested within nest as random intercept. Use log-transformed winsorised latency and square-root-transformed proportion of time persisting as responses. 6. Mother–offspring resemblance. For each trait (latency, persistence, improvement), calculate one per-individual mean collapsed across all trials and sessions. Average offspring values at the nest level. Fit simple linear models with maternal performance as predictor and mean offspring performance as response. Log-transform latency values; square-root-transform offspring improvement scores. 7. Model inference. Obtain p-values using Type II Wald F-tests with Kenward–Roger degrees of freedom (car::Anova). Full models including barrier type and order are available from the corresponding author on request.
Institutions
- University of AntwerpFlanders, Antwerp
- Ghent UniversityFlanders, Ghent
- Museo Nacional de Ciencias NaturalesMadrid, Madrid