By embedding subgroups in a sociogram w e generate images consistent with long standing theoretical descriptions of network structure. Moreover, these images sustain an eclectic array of theoretically based interpretations, thus providing a basis for theoretical integration. We generate and interpret two images, one of professional discussions among high school teachers and the other of friendships among the French financial elite. Our findings su ggest that this approach has great potential for informing characterizations and interpretations of the structure of various social networks.

Social networkers frequently make use of drawings to communicate information and ideas about networks. However, the impact of the layout of a network on the conclusions that a viewer is likely to draw has so far received very little scrutiny. In this paper, we extend work begun in (Blythe et al, 1995) and (Mcgrath et al, 19961 to understand how the layout of graphs depicting social network data influences the inferences viewers draw about social networks. Our previous work focused on the perception of prominence or bridging of a particular node. Here we focus on perceptions of clustering among nodes.

Network exchange theories predict how profits from exchange are divided among network positions. The network structure, or pattern of exchange relations, is the initial condition upon which predictions are based. Previous research has been limited by the small number of networks found suitable to test theories. We propose to accelerate theoretical development by automating analysis of network structures to find strategic test sites. To do so, two prerequisites must be met: (a) two successful theories of network exchange must be sufficiently formal to incorporate into a computer program and (b) a program must be developed to systematically generate network structures . We describe recent developments in network exchange theory that meet both prerequisites. Not only do we now have two successful theories of network exchange incorporated into computer programs, but at least two successful computer models of network exchange also exist. In addition, a prototype computer program has been developed to systematically generate all exchange networks with a given number of positions. Having met the prerequisites of automated analysis, we can now build a master program which generates networks, produces predictions from two or more theories for the power of positions in those networks, and compares predictions. Strategic test sites are indicated when predictions of two theories differ. Automated theoretical analysis has the potential to accomplish in a few hours what had required years of patience, skill, and luck. Once a strategic site has been located, experiments using human subjects can be conducted and other empirical evidence can be collected to aid theoretical development. Then, automated analysis can be re-applied using the developed theories on ever more complex networks . We may soon be able to analyze networks of the size and complexity typically found in society. Possibilities for automated theoretical analysis exist in other areas of sociology as well.

By embedding subgroups in a sociogram we generate images consistent with longstanding theoretical descriptions of network structure. Moreover, these images sustain an eclectic array of theoretically based interpretations, thus providing a basis for theoretical integration. We generate and interpret two images, one of professional discussions among high school teachers and the other of friendships among the French financial elite. Our findings suggest that this approach has great potential for informing characterizations and interpretations of the structure of various social networks.