providing a soft and shadow-free lighting. EthoVision was set to detect animals by means of background subtraction, and the parameters of size and contrast were set in pilot tests to optimize object detection.

Results and Discussion We recorded the time spent in each of the four arms for every individual. In previous studies we had verified that when the animals were in the arms, they were mainly engaged in feeding and interacting with the mirror [5]. The time spent in the arm was therefore used as an approximation for their behavior. The preliminary results indicate a phenotypic correlation between growth and feeding patterns, and perhaps between growth and social tendency. QTL analyses are presently carried out. Preliminary data indicate a genetic influence on the behavior, which may perhaps also throw light on the genetic connection between production traits and observed behavior. The results may be helpful for designing breeding programs which will allow animals to maintain more of their natural behavior while still producing at an efficient level. Our experiments also show that EthoVision can be used successfully to automate behavioral observations in fowl.