COMPLEXITY, THRESHOLDS AND CATEGORIES IN NEIGHBORHOOD RESEARCHThe majority of theoretical questions in neighborhood research center on categorical or interval constructs—phenomena that at least implicitly have distinct interval thresholds and/or boundaries between classes. Such constructs include temporally cross-sectional phenomena such as urban blight, ethnic or other subpopulation enclaves, territories of the underclass, and local niches of particular economic activities as well as dynamic processes such as gentrification, abandonment, and group transitions. The problem this presents to urban scholars is to set boundaries for these intervals and categories when the measures used are predominantly continuous. Formally, the challenge is to properly define and oper-ationalize thresholds: to correctly convert scales of measurement from continuous to ordi-nal or interval, in ways that conform to theory. Even the simplest case, in which boundaries must be drawn for the values of a single variable, is in fact far from simple. Because urban reality is messy, however, combinations of characteristics often define constructs and functional areas, so that the category or interval thresholds must be defined in multivariate data space. This is not a major obstacle to simple threshold treatment as long as the set of variables is a vector and as long as the outliers in multivariate space are not theoretically interesting. For example, one may reasonably construct a weighted index of urban blight from the number or density of broken windows, graffiti, and abandoned cars; this index can be collapsed into intervals of the degree of blight. But what if the theo-retically meaningful categories are complex multivariate permutations? What if we need to empirically distinguish wealthy Black suburbs from both wealthy White suburbs and moderate-income Black suburbs? Such situations require multivariate classification tech-niques rather than a vector of continuous variables or a single, continuous index statistic to distinguish meaningful types of variation among local areas. The evolution of cities in late capitalism has contributed greatly to the complexity of urban processes and phenomena in ways that make multivariate category and interval per-mutations in variation, and hence the need for classification techniques, more important than before. Whereas enclave formation for a given group was a simple, one-directional proposition in the 1920s and 1030s, a group’s enclaves in a single city may now be vari-ously growing, disappearing, or changing hands today. Whereas the rise of suburbs in the 1950s was a straightforward process of new low-density peripheral development outside