Abstract: Flocking is a paradigmatic example of self-organized collective behaviour, where collective ordering emerges from the mutual interactions between  individuals. In this respect, it shares striking similarities with collective phenomena in inanimate systems that have long been studied by the physics community. Still, biological systems are more complicated than physical ones and it is not evident whether they can be described by the same kind of general laws so well understood in physics. Experimental findings often go beyond simple expectations, making this field even more fascinating. In this talk I will discuss our attempts to study collective animal behaviour using  a physicist's perspective. I will show how we used concepts and methods from statistical physics to make sense of experimental data of large flocks of birds. In particular, I will focus on the velocity correlation functions, that well capture the balance between consensus and independence among the large number of individuals in the group. We measured these correlations in real flocks of starlings and found that they exhibit a non-trivial scale-free behavior, indicating a surprisingly large degree of coordination and collective response. I will describe how we can use these correlations to systematically address the inverse problem, extract from the experimental measurements information on the underlying microscopic interactions, and build minimal (maximum entropy) models directly from the data.