1. Floods are major disturbances to stream ecosystems that can kill or displace organisms and modify habitats. Many studies have reported changes in fish assemblages after a single flood, but few studies have evaluated the importance of timing and intensity of floods on long-term fish assemblage dynamics. 2. We used a 10-year dataset to evaluate the effects of floods on fishes in Kings Creek, an intermittent prairie stream in north-eastern, Kansas, U.S.A. Samples were collected seasonally at two perennial headwater sites (1995-2005) and one perennial downstream flowing site (1997-2005) allowing us to evaluate the effects of floods at different locations within a watershed. In addition, four surveys during 2003 and 2004 sampled 3-5 km of stream between the long-term study sites to evaluate the use of intermittent reaches of this stream. 3. Because of higher discharge and bed scouring at the downstream site, we predicted that the fish assemblage would have lowered species richness and abundance following floods. In contrast, we expected increased species richness and abundance at headwater sites because floods increase stream connectivity and create the potential for colonisation from downstream reaches. 4. Akaike Information Criteria (AIC) was used to select among candidate regression models that predicted species richness and abundance based on Julian date, time since floods, season and physical habitat at each site. At the downstream site, AIC weightings suggested Julian date was the best predictor of fish assemblage structure, but no model explained >16% of the variation in species richness or community structure. Variation explained by Julian date was primarily attributed to a long-term pattern of declining abundance of common species. At the headwater sites, there was not a single candidate model selected to predict total species abundance and assemblage structure. AIC weightings suggested variation in assemblage structure was associated with either Julian date or local habitat characteristics. 5. Fishes rapidly colonised isolated or dry habitats following floods. This was evidenced by the occurrence of fishes in intermittent reaches and the positive association between maximum daily discharge and colonisation events at both headwater sites. 6. Our study suggests floods allow dispersal into intermittent habitats with little or no downstream displacement of fishes. Movement of fishes among habitats during flooding highlights the importance of maintaining connectivity of stream networks of low to medium order prairie streams. ?? 2006 The Authors.