Social interaction patterns are relevant to explain (social) travel behavior. As such, the objective of this paper is to comparatively study the factors that influence social interaction frequency among social network members with different communication modes. Based on data from seven surveys on social networks, this analysis seeks to shed some light on (i) the similarities and differences in social interaction frequency patterns, (ii) the relation of personal and network characteristics with observed patterns, and (iii) the extent to which these associations are consistent across contexts, in terms of effect direction and magnitude. A multilevel-multivariate lognormal hurdle model is used to jointly analyze social interaction frequency patterns across all datasets. Level 1 includes information on ego-alter dyad characteristics, level 2 includes ego level socio-demographic and aggregate social network characteristics, while level 3 includes information specific to each context where data was collected. In line with network capital theory, results show the existence of very consistent associations between social interaction frequency and some network and dyad characteristics such as network size, ego-alter distance, and emotional closeness, which showed some degree of generality irrespective of context. Building up on previous research, results also suggest that the effect of a higher transport cost-to earnings ratio is more likely to manifest in the tie-formation phase, in such a way that the geographical spread of the network will tend to be smaller, but conditional on such a network distribution, the cost-to-earnings ratio effect becomes negligible. For other variables such as education level, gender and relationship type, effect patterns were less clear, which might be explained by socio-economic, and other contextual factors, as well as methodological differences across studies. The model presented here can provide average levels of demand for social interactions, which bounded by the geographical distribution of networks, can be used to further understand travel demand in urban environments and transportation systems at the local or regional level.