Terrestrial runoff represents a major source of mercury (Hg) to aquatic ecosystems. In boreal forest catchments, such as the one in northern Sweden studied here, mercury bound to natural organic matter (NOM) represents a large fraction of mercury in the runoff. We present a method to measure Hg stable isotope signatures of colloidal Hg, mainly complexed by high molecular weight or colloidal natural organic matter (NOM) in natural waters based on pre-enrichment by ultrafiltration, followed by freeze-drying and combustion. We report that Hg associated with high molecular weight NOM in the boreal forest runoff has very similar Hg isotope signatures as compared to the organic soil horizons of the catchment area. The mass-independent fractionation (MIF) signatures (?199Hg and ?200Hg) measured in soils and runoff were in agreement with typical values reported for atmospheric gaseous elemental mercury (Hg0) and distinctly different from reported Hg isotope signatures in precipitation. We therefore suggest that most Hg in the boreal terrestrial ecosystem originated from the deposition of Hg0 through foliar uptake rather than precipitation. Using a mixing model we calculated the contribution of soil horizons to the Hg in the runoff. At moderate to high flow runoff conditions, that prevailed during sampling, the uppermost part of the organic horizon (Oe/He) contributed 50-70% of the Hg in the runoff, while the underlying more humified organic Oa/Ha and the mineral soil horizons displayed a lower mobility of Hg. The good agreement of the Hg isotope results with other source tracing approaches using radiocarbon signatures and Hg?:?C ratios provides additional support for the strong coupling between Hg and NOM. The exploratory results from this study illustrate the potential of Hg stable isotopes to trace the source of Hg from atmospheric deposition through the terrestrial ecosystem to soil runoff, and provide a basis for more in-depth studies investigating the mobility of Hg in terrestrial ecosystems using Hg isotope signatures.