Experimental stress does not increase fluctuating asymmetry of broiler chickens at slaughter age

Fluctuating asymmetry (FA) is increasingly applied as a putative indicator of animal welfare. Yet its sensitivity to measure welfare of ad libitum-fed farm animals (that presumably have little or no energy allocation constraints) remains largely untested. This study was conducted to examine whether FA is sensitive to experimentally induced stress in broiler chickens and whether effect sizes differ between emotional and physical stressors. Broiler chickens were randomly assigned to emotional stress treatments (pain or frustration), physical stress treatments (wet litter or high temperature and density), or no stress treatment (control). Both physical stressors, unlike the emotional stressors, were known to affect a number of conventional welfare indicators measured at slaughter age. Left-right asymmetry of 14 bilateral traits was measured at slaughter age and compared between treatments. Seven of the 14 bilateral traits proved unsuitable for the study of FA, either due to the presence of directional asymmetry or high measurement error. Fluctuating asymmetry tended to be lowest in the control group and highest in the high temperature and density treatment. However, either when modeling traits as repeated measures at individual broiler level or when performing trait-by-trait analysis, no significant differences between treatments were detected. This negative result may indicate that FA is not a suitable indicator to detect variations of welfare status in fast-growing broiler chickens because of strong past selection for increased BW and improved feed efficiency, which can mask additional stress effects on developmental processes. Alternatively, FA is not a sensitive indicator of welfare in ad libitum-fed animals because of absence of energy allocation constraints. Finally, FA may still be a suitable indicator of welfare under such conditions, but differences between treatments may remain undetected due to insufficient statistical power, which was estimated at 35% for our study.