Increased somatic growth rate can reduce the concentration of persistent toxins such as mercury in fast-growing relative to slow-growing individuals, a process called growth dilution.  However, it is not clear to what extent growth dilution explains variation in toxin concentrations in the field.  We used a unique, large-scale field assay, releasing juvenile Atlantic salmon from uniform initial conditions into natural streams, to evaluate the relative importance of growth dilution and prey contamination in determining mercury concentrations ([Hg]) of stream-dwelling fish (18 sites, repeated over two years).  Salmon rapidly accumulated mercury (mean 1400% increase in [Hg] over one summer), but accumulation rates were highly spatially variable (8-10 fold range in site means).  Mean salmon [Hg] generally tracked prey [Hg] across sites.  However, salmon growth accounted for 40% of the explained variation in average salmon [Hg] independent of prey [Hg].  Consistent with growth dilution, faster-growing salmon had lower [Hg], both on average across sites and across individuals within sites.  Further analysis with a mass-balance model shows that growth dilution is a key driver of fish [Hg] when prey [Hg] is high and growth efficiency drives variation in fish growth.