In studies of epithermal precious and base metal ore deposits, estimates of salinity (total dissolved salts) are frequently in error when based on fluid inclusion ice melting measurements in the absence of an independent determination of the CO 2 content of the inclusion fluid. For a fluid of known composition the melting pint of ice (T m ) may be calculated from T m = -sigma K i m i where K i is the molal freezing (or melting) point depression constant and m i is the molality of a component i (i = Na (super +) , K (super +) , Cl (super -) , CO 2 etc) for fully dissociated solute species such as Cl (super -) , in the salinity range considered here K = 1.72 Kelvin/molal, and for undissociated, nonpolar species such as CO 2 , K = 1.86 Kelvin/molal. Fluid inclusion ice-melting data from New Zealand geothermal fields correlate well with values calculated using the above equation and the measured compositions of discharges from wells from which the inclusion samples were obtained. Loss of the dominant dissolved gas, CO 2 during boiling at depth results in large, systematic decreases in apparent salinity (in terms of T m ) in the Broadlands field. Misinterpretation of fluid inclusion freezing data may lead to substantial errors in the reconstruction of the physico-chemical environment of ore formation in fossil systems. For example in the absence of CO 2 analyses, inclusion fluids similar in gas content to the Broadlands geothermal fluid (NaCl 0.2 wt %, CO 2 up to 4.4 wt %) may be interpreted to have salinities of 0.85 wt percent NaCl leading to errors of 23% in the estimated depth of formation at 280 degrees C -0.5 unit in the estimated pH of the ore fluid, and on the order of 200 times in the estimated solubility of an ore component such as lead. Such errors may be transmitted into subsequent estimation of fluid flux or duration of ore formation.