Abstract The analysis of defensive or fortified archaeological sites in the Pacific has a long history, with Geographic Information Systems (GIS) approaches applied with increasing frequency. Much of the recent GIS-based research has emphasized views to and from defensive sites and site location relative to resources such as agricultural land. We add to this growing body of research with analyses of defensive sites in the western islands of Fiji. Our work is the first quantitative GIS analysis of visibility in the Pacific and examines views to and from sites and the content of those views with statistical comparisons to random background samples. Our results indicate that views of high fertility land were an important consideration in the placement of some defensive sites, but that views of other defensive sites were not important. Additionally, some sites are deliberately placed in areas to obscure their visibility. The defensive sites of Remote Oceania are abundantly distributed across island landscapes from the artificially sculpted hilltops of New Zealand (Sutton et al. 2003), to the ring-ditched villages of Fiji’s deltas (Parry 1982), and the forts atop the peaks of French Polynesian islands (Kennett et al. 2006). Variation in defensive site features and locations has for some time been explained as a result of a site’s relative proximity to and control of arable land (e.g. Kirch 1984). More recently archaeologists have built upon earlier research and generated quantitative analyses of agricultural potential, palaeoenvironments, and other relevant variables to explain the location of defensive sites (e.g. Field 2008; Kennett et al. 2006). Visibility from and between defensive sites has also been considered important in explaining site location (e.g. Field 1998; Parry 1977). We build upon this research on site visibility and defence with the first quantitative geographic information systems (GIS) analyses to test whether visibility was an important consideration in the location of prehistoric sites in western Fiji. We investigated 34 sites with defensive architecture in upland and relatively inaccessible locations identified previously in the Yasawa and Mamanuca Islands of western Fiji (Cochrane 2009; Cochrane et al. 2007, 2011; Hunt et al. 1999) (Figure 1). We first examined the size of views from these sites, the content of those views, and the size of the area within which the site could be seen. We then compared these observed characteristics of defensive and upland sites to a constrained random sample of other locations on the islands. Constrained random sample locations, or back- ground samples, are generated from (i.e. constrained by) areas geographically similar to the known defensive and Archaeol. Oceania 46 (2011) 76–84 76 How is visibility important for defence? A GIS analysis of sites in the western Fijian Islands CECILIA SMITH and ETHAN E. COCHRANE Keywords: defensive sites, GIS, viewshed, Kolmogorov-Smirnov test, agriculture CS: Department of Anthropology (M/C 027), University of Illinois at Chicago, 1007 W. Harrison Street, Chicago, IL 60614, USA. csmith83@uic.edu. EEC: International Archaeological Research Institute, Inc., 2081 Young Street, Honolulu, Hawai‘i, 96826, USA. ecochrane@iarii.org Figure 1. Western Fiji showing sites of the Yasawa and Mamanuca Islands in the analysis indicated by grey boxes. Individual island names are italicised. Individual site locations on Naviti not shown at this scale for clarity.