We study the extent of segregation in the social space of urban America. We measure segregation as the (lack of) actual personal connections between neighbourhoods as opposed to conventional measures that assume the strength of these connections. We distinguish social segregation from geographical definitions of segregation, building and comparing city-level indices of each. We apply our measures to the seventy-five largest MSAs in the USA. Cities like Miami, Washington DC, and Cincinnati rank higher in social segregation than they do based on the conventional residential isolation, while New Orleans, San Francisco, and Richmond fall in ranks. Conditional on residential segregation, cities with more institutions that foster social cohesion (churches and community associations) are less socially segregated. We also decompose social segregation into residential, within-neighbourhood, and cross-neighbourhood interaction components, revealing that while social ties mirror residential patterns, geographically broader networks can partially offset residential isolation. Examining heterogeneity across racial minority groups, we find variation in social isolation with Black individuals being relatively more socially segregated than Hispanic and Asian minorities. Our results suggest that social connections, beyond residential geography, are important for understanding racial segregation in America.