Raman spectroscopy has been used to obtain the concentration of H2S in individual fluid inclusions in the Archean Missouri gold deposit in Western Australia. These results indicated that the gas phase of Type III, vapor-rich fluid inclusions from the Missouri gold deposit contained highly variable CO2/CH4 ratios and mole fractions of H2S up to 0.0018. There is good agreement between the calculated and empirically estimated H2S contents at 300°C and 200 MPa. The measured H2S concentrations can be explained by the proposed equilibrium with pyrite in the vicinity of or along the pyrite-pyrrhotite sulfide buffer. In this study, optical methods were used for estimating the volume fractions of vapor and liquid (where observable) in inclusions at room temperature. To investigate the uncertainty associated with these measurements, where no aqueous phase was observed in an inclusion, we have calculated its bulk composition both without liquid water and also by assuming that there was a 0.5 μm film of liquid water on the walls of the inclusion. Calculations of bulk densities were found to be considerably higher for inclusions with liquid water when compared with vapor-only inclusions. To investigate the error associated with undetected water in vapor-rich inclusions, numerical experiments were completed for a number of fluid compositions in the H2O-CH4-CO2-H2S compositional space, both for a regular compositional grid and for a set of specific fluid compositions derived from the visual observations and Raman measurements. Using this approach, the bulk compositions of fluid inclusions were calculated in equilibrium with the pyrite-pyrrhotite buffer within the temperature-pressure range of 270 to 360°C and approximately 40 to 260 MPa. The numerical modeling has also shown that variations in calculated H2S concentrations of up to two orders of magnitude may arise due to errors associated with visual estimates of volume fractions in vapor-rich inclusions. To reduce this source of error, bulk inclusion compositions should be checked by other analytical methods, where possible.