To determine the speciation and concentrations of dissolved COH volatiles in graphite-saturated martian primitive magmas, we conducted piston-cylinder experiments on graphite-encapsulated synthetic melt of Adirondack-class Humphrey basaltic composition. Experiments were performed over three orders of magnitude in oxygen fugacity (IW+2.3 to IW-0.8), and at pressures (1-3.2GPa) and temperatures (1340-1617°C) similar to those of possible martian source regions. Oxygen fugacities were determined from compositions of coexisting silicate melt+FePt alloy, olivine+pyroxene+FePt alloy, or melt+FeC liquid. Infrared spectra of quenched glasses all show carbonate absorptions at 1430 and 1520cm-1, with CO2 concentrations diminishing under more reduced conditions, from 0.50wt% down to 26ppm. Carbon contents of silicate glasses and FeC liquids were measured using secondary ion mass spectrometry (SIMS) yielding 36-716ppm and 6.71-7.03wt%, respectively. Fourier transform infrared (FTIR) and SIMS analysis produced similar H2O contents of 0.26-0.85 and 0.29-0.40wt%, respectively. Raman spectra of glasses reveal evidence for OH- ions, but no indication of methane-related species. FTIR-measured concentrations of dissolved carbonate diminish linearly with oxygen fugacity, but more reduced conditions yield greater dissolved carbonate concentrations than would be expected based on oxidized conditions in previous work. C contents of silicate glasses determined by SIMS are consistently higher than C as carbonate determined by FTIR. Their difference, termed non-carbonate C, correlates well with additional IR absorptions found in reduced glasses (fO2<IW+0.4) at 1615, 2205, and 3370cm-1. These absorption bands are not seen in more oxidized glasses, except B441 (IW+1.7), presumably because they represent reduced C-bearing complexes. The 2205cm-1 peak is attributed to a CO complex, possibly an Fe-carbonyl ion. The 1615cm-1 peak does not correlate with that at 2205cm-1, but does correlate with non-carbonate C and is in a region commonly associated with CO bonding. The origin of the peak at 3370cm-1 is poorly understood and could potentially be owing to a variety of COH species or to NH bonding. The intensities of the 1615 and 3370cm-1 peaks correlate with each other leading us to provisionally attribute both to an unspecified complex with both CO and NH bonds. These results suggest that dissolved species such as carbonyl or other CO-bearing species could be a significant source of C fluxes to the martian atmosphere, with minor additions of CO2 and negligible methane contributions. By assuming that degassed, reduced C ultimately becomes atmospheric CO2, reduced C outgassing may be incorporated in models of martian atmospheric evolution. At Humphrey source region conditions (1350±50°C, 1.2±0.1GPa) the total C contents are equivalent to 1200ppm CO2 at IW+1 and 475ppm CO2 at IW, which are 2 and 4 times higher than the CO2 derived from CO32- alone. For reasonable magmatic fluxes over the last 4.5Ga of martian history, such graphite-saturated magmas would produce 0.25 and 0.60bars from sources at IW and IW+1, significantly more than expected solely from consideration of dissolved CO2. The carbon contents of FeC liquids in this study are consistent with graphite-saturated carbide liquids becoming more C-rich with increasing temperature. Experiments with melt and FeC liquid have values of DCall/sil between 1.3×103 and 2.2×103, potentially allowing planetary mantles to retain significant C following core formation.