Social Networks 34 (2012) 47–58 Contents lists available at ScienceDirect Social Networks journa l h o me page: www.elsevier.com/locate/socnet Social networks and spatial configuration—How office layouts drive social interaction Kerstin Sailer a,∗ , Ian McCulloh b a The Bartlett School of Graduate Studies, University College London, 1-19 Torrington Place, London WC1E 7HB, United Kingdom b United States Military Academy, Network Science Center, 601 Cullum Road, West Point, NY 10996, United States a r t i c l e i n f o Keywords: Intra-organizational networks Interaction patterns Exponential random graph model Office layout Spatial configuration Space Syntax a b s t r a c t This paper analyzes the spatial dimensions of office layouts in diverse knowledge-intensive workplace environments based on the theoretical and methodological propositions of Space Syntax, and brings this together with the analysis of intra-organizational interaction networks. Physical distances between agents are modeled in different ways and used as explanatory variables in exponential random graph modeling. The paper shows that spatial configuration in offices can be considered an important but not sole rationale for tie formation. Furthermore, it is shown that spatial distance measures based on detailed configurational analysis outperform simple Euclidean distance metrics in predicting social ties. © 2011 Elsevier B.V. All rights reserved. 1. Introduction Beginning with proclamations by the sociologist Daniel Bell (1976) and later coined as ‘spatial turn’, ‘renaissance of space’ (Maresch and Werber, 2002) or the ‘spatialization of social the- ory’ (Massey, 1998), the concept of space has regained interest and attention throughout many social sciences, the humanities, arts, philosophy and cultural theory. The cultural geographer Edward Soja proposed that “the spatial dimension of our lives has never been of greater practical and political relevance than it is today.” (Soja, 1996, p. 1) This has led to a growing research field consider- ing various aspects of space in studies of social phenomena. Besides conceptual and theoretical contributions, it has been demonstrated empirically that space matters; for instance it has been argued that failing to control for the effects of spatial autocorrelation in regres- sion models in cases with a spatial bearing can bias the estimation of other factors (Doreian, 1980, 1981; Dow et al., 1984; Mencken and Barnett, 1999). Likewise, scholars have integrated geographical perspectives into the investigation of networks and the rationales for tie for- mation. They have conceptualized the spatiality of networks from a more theoretical point of view, as a combination of highly local- ized clusters with some additional random links to create global short path length in networks. This has subsequently been found to be a common structural feature of many real world networks by Watts (2004). These findings mean that location, or proxim- ∗ Corresponding author. Tel.: +44 20 3108 3162; fax: +44 20 7916 1887. E-mail address: k.sailer@ucl.ac.uk (K. Sailer). ity between agents have become an interesting aspect to study in networks, since proximity gives rise to clustering, which in turn is seen as a main structural component of networks. Of course prox- imity in this sense is a broad term and is not confined to physical space. Localized clusters may also derive from structural considera- tions like reciprocity and transitivity (Holland and Leinhardt, 1972; Wasserman and Faust, 1994), or social similarity between agents, i.e. homophily (Blau, 1977; Ibarra, 1992; McPherson et al., 2001). Still space and physical proximity may play a crucial role in sup- porting tie formation and localized clusters and as such are factors worth investigating more closely. This paper therefore sets out to explore how the study of physi- cal space can prove a fruitful endeavor in network related research. An approach for including spatial data in exponential random graph models (ergm) as edge covariates is proposed. Four competing methods of measuring space and their effects on social interaction are compared with each other and against simple Euclidean dis- tance, focusing on spatial micro settings within office buildings. The significance of the spatial effects in network ergm models suggests the importance of controlling for spatial distance in network mod- els in corollary to the work on spatial autocorrelation in multiple linear regression. Previous studies of organizational behavior and communication patterns within organizations have argued that physical proxim- ity plays a major role for the probability of communication. In a seminal study by Allen and Fustfeld (1975) it was shown that co- workers that were separated by more than 25 m walking distance had a significantly lower probability of communicating with each other than those co-located closer to each other. More recently these results were confirmed by showing that highly frequent interaction (on a daily basis) did not reach further than 18 m on 0378-8733/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.socnet.2011.05.005