As ecological patterns are scale dependent, making decisions about sampling design critical. In the context of community assembly, many metrics have been developed to quantify species segregation and aggregation, of which the checkerboard metric (C-score) and the V-ratio are widely used. Using data that describe the spatial pattern of four species-rich, Mediterranean-climate, shrubland communities, we confirm the scale dependence of these metrics. At three of the four sites, we observed that the co-occurrence metrics monotonically increased across all the spatial grains we assessed; at the other site, the previously reported hump-shaped relationship between spatial grain and co-occurrence metrics was observed. The spatial grain at which measures of co-occurrence peaked varied across the plant communities and taxa and increased with exclusion of the less frequently sampled species. Inferences about community assembly using co-occurrence metrics must recognise their inherent scale dependence, and the inferential tools developed in spatial point pattern analysis may facilitate this task.