AbstractHabitat fragmentation in aquatic systems has led to widespread isolation of stream fishes. Metapopulation theory predicts that persistence is directly related to local patch size and its characteristics, but because these relationships tend to be taxon‐specific, empirical data are important. We assembled 246 observations of occurrence of westslope cutthroat trout (WCT), a taxon of concern in the western U.S. and Canada, in stream networks isolated for up to 100 years (median 40 years) above human‐made barriers, mostly culverts, at road crossings within U.S. National Forests. We used logistic regression to analyse how WCT occurrence varied with patch size, isolation time and stream‐level covariates. Occurrence was positively related to stream length and habitat quality within the isolated stream network and negatively related to elevation and channel gradient. Unexpectedly, the probability of occurrence was not related to how long a habitat patch had been isolated. At the median elevation (1354 m) and channel gradient (14%), and where habitat quality was poor, WCT were likely to occur (probability >0.5) if an isolated stream network was at least 1.7 km. If habitat quality was high, about 0.2 km of habitat produced the same probability. Although there are important limitations, this analysis provides the first empirical estimate for how patch size and patch‐level characteristics influence persistence of WCT in isolated stream networks.