1. We hypothesised that adult insects actively monitor potential habitats for the presence of fish by means of chemical cues and avoid sites that pose significant risks. This was examined by quantifying colonisation of insects in outdoor pools with no fish (controls), fish (direct predation effect) or caged fish (chemical predator cues). 2. A significant direct effect of predation was found, but no indirect effect (avoidance of chemical cue pools), on the total biomass of colonising insects. However, predatory insects avoided fish-cue pools, thus releasing non-predatory insects from predation. This resulted in significantly greater biomass of non-predatory insects in fish-cue pools than control pools. 3. Fish reduced the number of species of colonising insects in pools through predation. This negative influence of fish implies that caution is necessary when stocking wetlands and ponds with fish if the goal is to maximise biodiversity. 4. Our data suggest that although predatory aquatic insects may use chemical signals to assess the quality of potential habitats with respect to predation risk, direct predation is the main method by which fish affect insect assemblages in ponds. Because fish and invertebrate predators may both have strong effects on prey mortality, behavioural adjustment by insects to the actual predator regime within a habitat should be more important than avoiding colonisation of habitats with fish.