Many animal aggregations display collective patterns on the large scale, ultimately due to the interactions between the individuals in the group. Recent findings on flocks of birds and swarms of insects show that these groups exhibit strong mutual correlations and quick mechanisms of information propagation, signatures of the efficient collective response to external perturbations. Besides, they obey static and dynamic scaling laws suggesting that we can use a statistical physics approach to describe the large scale, and define novel `classes' of behavior. I will review our current understanding of collective animal behavior and discuss how a physics based perspective, from experiments to modelling, can help to define a unified description for these systems.