Abstract. 
New technologies give new ways of studying social and
behavioral networks thanks to unprecedented access to different types
of data sets.  In this talk, I will first present recent work on a
detailed empirical study of an online social network, showing the
presence of strong heterogeneities between actors, and how the
presence of homophily can be measured. I will also show how
longitudinal studies allow us to detect the simultaneous presence of
selection and influence effects. I will then briefly present how it is
possible to measure behavioral networks of face-to-face proximity
thanks to a newly developed infrastructure. Such networks can be
measured in different contexts such as schools, hospitals, scientific
conferences.  Interestingly, it can then become possible to compare
the behavior of conference attendees with their online social
relations.  We find that the existence of an on-line link between
individuals is strongly correlated with the strength of their
face-to-face presence, as well as with the similarity of their social
contacts in physical space. Based on these findings, we characterize
how accurately the existence of an on-line social link can be
predicted by using the measured properties of face-to-face presence.