According to the restricted movement paradigm (RMP), salmonid fry should typically disperse less than 50 m upstream and 500 m downstream from nests or stocking locations. However, recent evidence of juveniles dispersing up to 1600 m suggests that the RMP may be oversimplified for juvenile Atlantic salmon. To test the RMP, we implanted 679 997 eyed Atlantic salmon eggs (Salmo salar L.) into 19 artificial redds over six years in seven tributaries of a large river and then recovered fry via electrofishing to characterize their dispersal over their first summer of life. As expected, most fry dispersed downstream, but an average of 35% of fry moved upstream. Surprisingly, fry moved just as far upstream as downstream (medians = 403 and 404 m, respectively), with average maximum distances of 1.23 km and 2.14 km, respectively. Fry were larger at lower densities and farther from redd sites, consistent with density-dependent growth. After controlling for density, however, fry were larger upstream than downstream, suggesting that larger fry move upstream compared to downstream. While there was variation among streams and years, kurtosis values were largely consistent with a normal distribution around the redd site, with only two of the 19 distributions being leptokurtic. Tributaries had even mixtures of mobile and stationary individuals, indicating a largely homogeneous movement strategy. Our data suggests that salmon fry were more mobile than previously thought, which should facilitate their stocking or reintroduction to new habitats.