Diamond [Assembly of species communities. In: Ecology and Evolution of Communities (eds Cody, M.L. & Diamond, J.M.). Harvard University Press, Cambridge, pp. 342-444] proposed that resource competition leads to checkerboard-like distributions of competing species. This proposal prompted research that revealed checkerboard patterns within a wide range of communities, but the mechanisms that generate such patterns are still poorly understood. Here we present whole-lake natural experiments and analyses of species-environment relationships in small coastal lake fish communities that were fragmented when land uplift isolated these lakes from the Baltic Sea, showing that a combination of predation and habitat suitability generated checkerboard distributions. Checkerboard patterns developed because two piscivores, northern pike and Eurasian perch, caused the extinction of several prey species in deep lakes. Conversely, low oxygen levels in shallow lakes caused extinction of the piscivores, and these areas served as a refuge for tolerant prey species. Based on these findings, we suggest that habitat suitability and biotic interactions should be viewed simultaneously in null models of assembly rules.

1. The correlation between water chemistry, physical variables and fish community composition was examined in 40 small (£30 ha) coastal lakes in northern Sweden. Twenty of the 40 lakes were isolated from other water bodies and 20 were connected to the Baltic Sea. Lakes were fished in summer, using three different methods. Water chemistry was sampled in late winter prior to ice-out and pH was measured additionally in summer. 2. Our central question was whether water chemistry plays a greater role in the composition of fish communities in isolated lakes than in connected lakes, as isolated lakes cannot be recolonised once a species has become extinct. 3. Results indicate that winter anoxia affects community composition only in isolated lakes, whereas acidity is of importance in both connected and isolated lakes. Methane (indicating anoxia), was significantly correlated with variation in fish community composition in isolated lakes, and a group of variables that indicate anoxia (CH 4 , pCO 2 , inorganic carbon and dissolved oxygen) explained 24-34% of the variation. pH alone explained 12% of the variation in community composition for connected lakes and a group of variables indicating acidity (summer and winter pH and ANC) explained 10-20% of the variation in isolated lakes. Lake area was the most important physical variable, being significantly correlated with the variation in fish community composition in connected lakes. 4. In isolated lakes, the presence of pike (Esox lucius), perch (Perca fluviatilis) and roach (Rutilus rutilus) was associated with low CH 4 . The occurrence of crucian carp (Carassius carassius) and roach was positively correlated with pH, and the crucian carp was apparently also affected by predation by pike and perch. In connected lakes the effect of anoxia was low, probably due to the possibility of recolonisation and pockets of oxygenated water, allowing pike and perch to persist and thereby limiting the distribution of crucian carp.