How do social networks reveal and shape intergroup boundaries? Using the case study of friendship networks among adolescences in European schools, this dissertation addresses this central topic with three sequential questions: what drives the seemingly inevitable forces of network homophily? How do segregated networks evolve over time? What are the implications of peer networks for individuals’ identity profiles? The first chapter examines two factors through which process of network homophily could vary in shades. I first probe into the distinction between homophily as attribute selection (based on visible status markers) and as cultural matching (based on the perception of overall cultural profiles). I then propose a novel measure of contextual salience of ethnicity based on network modularity. Using cross-sectional Exponential Random Graph Models, my analysis of 584 classroom friendship networks reveals that process of attribute selection operates in tandem with processes of cultural matching. While both contribute to intergroup friendship segregation, the critical difference between the two is that attribute selection (e.g., preference for same ethnic peers) is sensitive to the contextual salience of group boundaries while cultural matching is not. While a lion’s share of scholarly attention is paid to network segregation, we know surprisingly little about what desegregates a network, or in Simmel’s (1955) term, what makes cross-cutting social circles. The second chapter takes a longitudinal perspective and asks what constitutes the secret recipe for the emergence of cross-cutting social circles over time. I synthesize two existing competing theses of network evolution –– (a) the relational amplification thesis, which emphasizes the structural endogeneity of network processes (such as reciprocity and triadic closure), and (b) the multi-dimensional homophily thesis, which underscores the self-limiting nature of homophily based on multiple attributes. Applying Stochastic Actor Oriented Models to two case studies –– changes in ethnic and gender segregation in two-wave school networks, I find that the combination of two particular factors — weak homophilous preferences and the competing relationship among multiple network mechanisms — provides the seed for network desegregation and facilitates the emergence of cross-cutting social circles. The third chapter delves into the implications of networks for individuals’ group identity profiles. Prior research has paid extensive attention to the composition of networks while overlooking the structural positions within networks. Filling this gap, this chapter examines the associations of both compositional diversity and structural brokerage with individuals’ identity profiles. Findings suggest a contrasting image for different groups: (1) for migrants, both compositional diversity and intergroup brokerage are associated with a stronger host country identification and a weaker ethnic identity, but intergroup brokerage plays a more robust role than compositional diversity; (2) for natives, compositional diversity is associated with a weaker national identity and plays a more robust role than intergroup brokerage.