Recent work in the sociology of taste has begun to grapple with the relational properties of traditional survey-based data using techniques inspired by network analysis. Despite productive results from this endeavor, critics rightly question the face and ecological validity of the vague macrogenre labels included in standard arts participation surveys (e.g., Classical, Rock, Rap) which feed into these novel methods. In this paper, I propose a link-clustering approach for discovering focused microgenres from standard survey-based information on cultural tastes, exploiting the underlying relational patterns realized by the indirect connectivity structure of genres (via people) in a two-mode network. The link-clustering approach partially answers two of the challenges of macrogenre critics: The fact that actual genres are overlapping and not crisply bounded and that there is hidden heterogeneity within the broad labels we usually focus on. To showcase the fruitfulness of the proposed approach, I engage in two “case studies” featuring the vague macro genres of “Heavy Metal” and “Latin/Salsa” music and show the focused microgenres produced by the link clustering procedure are segmented in ways that help resolve puzzles that have emerged in previous work.