HUTCHINS, EDWIN E.; PHD

                         THE UNIVERSITY OF WISCONSIN - MADISON, 1985

                         SOCIOLOGY, SOCIAL STRUCTURE AND DEVELOPMENT (0700)
 

                         This thesis reports an analysis of structure in children's friendship networks. It examines the relationship
                         between two classic facets or dimensions of network structure: clustering--the differentiation of actors
                         into network subgroups--and the ranking of actors along a status hierarchy. Numerous sociometric
                         studies have demonstrated that children's friendship networks exhibit significant tendencies toward both
                         clustering and ranking. Sociometric research tends to examine the various facets of network structure in
                         isolation, with scant attention paid to the relationships between them. The present research extends
                         sociometric tradition by empirically examining the relationship between clustering and ranking. The
                         relationship between clustering and ranking forms a continuum bounded by two ideal-type structural
                         models. The Ranked Clusters Model exhibits ranking between clusters, while in the Parallel Clusters
                         Model ranking exists only within clusters. A dyad-based approach to network analysis is used to define
                         operational measures of network structure. Multivariate regressions of network tie on dyad-level
                         structural variables yield measures of the cluster/rank relationship that control for other dimensions of
                         network structure. Analysis of data on children's friendship networks in 48 classrooms reveals substantial
                         variation across networks in the relationship between clusters and rank, with a mild overall tendency
                         toward parallel clusters. Variation between networks is in part determined by age, gender, and race. A
                         parallel-clusters structure is more likely among older children, in networks in which gender is a strong
                         predictor of network tie, and in racially mixed and predominantly black networks. This research indicates
                         that social network structure is multifaceted, and that its facets are interrelated. The relationships
                         between various facets of network structure have implications for the measurement of network structure
                         and the study of intergroup relations.

Social Systems Simulation Group P.O. Box 6904 San Diego, CA  92166-0904 Roland Werner, Principal Phone/FAX  (619) 660-1603