ABSTRACT. Geographic distances influence social connections inside cities – even in our digital age [1,2]. Both the perception of physical distance and the likelihood of forming social ties across locations are influenced by infrastructural elements, as previous studies showed: public transport increases social connectivity along routes [3]; barriers to physical mobility influence the creation of social connections [4]. To directly investigate the impact of highways on social connections, we analyse how the spatial configurations of a city’s social and infrastructural networks are correlated. We use a highly granular, georeferenced social network of mutual follower relationships between Twitter users in the top 50 US metropolitan areas [5]. For each of these cities, we create a gravity-law inspired configurational null model [6] that reflects the population density and distance distribution of users’ home locations. We then overlay the spatial network of social connections with the network of highways from OpenStreetMap, and measure the average number of highways crossed by social network ties. We find that the probability of an edge crossing at least one highway is significantly lower for real social connections than for the null model, and validate these findings by several multivariate regression models. Our results confirm that urban highways, apart from causing spatial segregation patterns, also have a directly measurable negative correlation with the density of social connections, and unveil the importance of infrastructure to the study of increasing inequalities and social network fragmentation in urban areas.