Early behavioral predictors of injurious pecking: exploration of behavioral profiles in two commercial strains of laying hens.

Injurious pecking remains one of the most persistent welfare and economic challenge in the egg industry. Despite decades of research, its multifactorial nature makes it hard to prevent and manage. In this longitudinal experimental study, we used an innovative approach to identify the behavioral profile of high peckers (P+) compared to low peckers (P-) in two strains of Lohmann laying hens: the Lohmann LSL classic and the Lohmann Brown classic (N = 90 per strain). These animals were tested in two test rounds at 5-10 weeks and 18-23 weeks of age to measure a panel of behavioral traits. Each round consisted of the following tests: detour, novel object, cognitive bias, multivariate and tonic immobility, aiming to characterize different aspects of birds' behavioral profile. Home pen behavioral observations were also performed to establish the time budget of the individuals. Between 31-38 weeks of age, pecking observations were performed to identify P+ and P- birds (6.5 hours per individual). Data were analyzed using (sparse) partial least square discriminant analysis ((s)PLS-DA) to identify variables that best characterized our two strains, and which best discriminate P+ from P- birds for each strain (20 birds pecking the most and the least were retained for the analysis). Results show that, overall, LSL birds were more active than Brown birds, which exhibited a more passive behavioral profile. For Brown birds, the model was able to classify birds as P+ or P- with an error rate of 26% and 35%, respectively. When considering the 18 most relevant variables, P+ birds appeared less fearful and more socially motivated than P-. For LSL birds, the model was able to classify birds as P+ or P- with an error rate of 30% and 48%, respectively, suggesting that these results should be interpreted with caution. These results highlight the complexity of predicting feather pecking and the importance of taking the strain into account when trying to identify reliable indicators.