We use a global 3-D model of atmospheric mercury (GEOS-Chem) to interpret worldwide observations of total gaseous mercury (TGM) and reactive gaseous mercury (RGM) in terms of the constraints they provide on the chemical cycling and deposition of mercury. Our simulation including a global mercury source of 7000 Mg yr-1 and a TGM lifetime of 0.8 years reproduces the magnitude and large-scale variability of TGM observations at land sites. However, it cannot capture observations of high TGM from ship cruises, implying a problem either in the measurements or in our fundamental understanding of mercury sources. Observed TGM seasonal variation at northern midlatitudes is consistent with a photochemical oxidation for Hg(0) partly balanced by photochemical reduction of Hg(II). Observations of increasing RGM with altitude imply a long lifetime of Hg(II) in the free troposphere. We find in the model that Hg(II) dominates over Hg(0) in the upper troposphere and stratosphere and that subsidence is the principal source of Hg(II) at remote surface sites. RGM observations at Okinawa Island (Japan) show large diurnal variability implying fast deposition, which we propose is due to RGM uptake by sea-salt aerosols. Observed mercury wet deposition fluxes in the United States show a maximum in the southeast, which we attribute to photochemical oxidation of the global Hg(0) pool. They also show a secondary maximum in the industrial Midwest due to regional emissions that is underestimated in the model, possibly because of excessive dry deposition relative to wet (dry deposition accounts for 68% of total mercury deposition in the United States in the model, but this is sensitive to the assumed phase of Hg(II)). We estimate that North American anthropogenic emissions contribute on average 20% to U.S. mercury deposition.