1. Colonisation and population recovery are crucial to species persistence in environmentally variable ecosystems, but are poorly understood processes. After documenting movement rates for several species of stream fish, we predicted that this variable would influence colonisation rates more strongly than local abundance, per cent occupancy, body size and taxonomic family. We also predicted that populations of species with higher movement rates would recover more rapidly than species with lower movement rates and that assemblage structure would change accordingly.

2. To test these predictions, we removed fishes from a headwater and a mainstem creek in southwest Virginia and monitored colonisation over a 2-year period. Using an information–theoretic approach, we evaluated the relative plausibility of 15 alternative models containing different combinations of our predictor variables. Our best-supported model contained movement rate and abundance and was 41 times more likely to account for observed patterns in colonisation rates than the next-best model. Movement rate and abundance were both positively related to colonisation rates and explained 88% of the variation in colonisation rates among species.

3. Population recovery, measured as the per cent of initial abundance restored, was also positively associated with movement rate. One species recovered within 3 months, most recovered within 2 years, but two species still had not recovered after 2 years. Despite high variation in recovery, the removal had only a slight impact on assemblage structure because species that were abundant in pre-removal samples were also abundant in post-removal samples.

4. The significance of interspecific variation in colonisation and recovery rates has been underappreciated because of the widely documented recovery of stream fish assemblages following fish kills and small-scale experimental defaunations. Our results indicate that recovery of the overall assemblage does not imply recovery of each component species. Populations of species that are rare and less mobile will recover more slowly and will be more vulnerable to extinction in systems where chemical spills, hydrological alteration, extreme droughts and other impacts are frequent.