An on-line Hg reduction technique using stannous chloride as the reductant was applied for accurate and precise mercury isotope ratio determinations by multi-collector (MC)-ICP/MS. Special attention has been paid to ensure optimal conditions (such as acquisition time and mercury concentration) allowing precision measurements good enough to be able to significantly detect the anticipated small differences in Hg isotope ratios in nature. Typically, internal precision was better than 0.002% (1 RSE) on all Hg ratios investigated as long as approximately 20 ng of Hg was measured with a 10-min acquisition time. Introducing higher amounts of mercury (50 ng Hg) improved the internal precision to <0.001%. Instrumental mass bias was corrected using 205Tl/203Tl correction coupled to a standard-sample bracketing approach. The large number of data acquired allowed us to validate the consistency of our measurements over a one-year period. On average, the short-term uncertainty determined by repeated runs of NIST SRM 1641d Hg standard during a single day was <0.006% (1 RSD) for all isotope pairs investigated (202Hg/198Hg, 202Hg/199Hg, 202Hg/200Hg, and 202Hg/201Hg). The precision fell to <0.01% if the long-term reproducibility, taken over 11 months (over 100 measurements), was considered. The extent of fractionation has been investigated in a series of sediments subject to various Hg sources from different locations worldwide. The ratio 202Hg/198Hg expressed as δ values (per mil deviations relative to NIST SRM 1641d Hg standard solution) displayed differences from +0.74 to −4.00‰. The magnitude of the Hg fractionation per amu was constant within one type of sample and did not exceed 1.00‰. Considering all results (the reproducibility of Hg standard solutions, reference sediment samples, and the examination of natural samples), the analytical error of our δ values for the overall method was within ±0.28‰ (1 SD), which was an order of magnitude lower than the extent of fractionation (4.74‰) observed in sediments. This study confirmed that analytical techniques have reached a level of long-term precision and accuracy that is sufficiently sensitive to detect even small differences in Hg isotope ratios that occur within one type of samples (e.g., between different sediments) and so far have unequivocally shown that Hg isotope ratios in sediments vary within approximately 5‰.