The Late Archean Witwatersrand Basin of South Africa represents the world's foremost gold deposit. The basin is also the erosional remnant of the world's oldest (2.02 Ga) and largest known impact structure, the Vredefort impact structure. The debate about the origin of the Witwatersrand gold resource—as a placer deposit, through influx of gold-bearing hydrothermal solutions into the basin or mobilization of basinal fluids, or in the form of a so-called modified placer deposit—has been ongoing for about a century. Currently, most active Witwatersrand workers favor either a hydrothermal process or the modified placer theory, although the nature and timing of hydrothermal events affecting the basin remains a matter of debate. Here, results of fluid inclusion petrographic and quadrupole mass spectrometric analyses on Ventersdorp Contact Reef conglomerate and footwall quartzite samples, as well as Vredefort impact-related pseudotachylitic breccias from the bedding-parallel fault zones at the top and bottom of the Ventersdorp Contact Reef, are reported. These analyses were obtained in an attempt to derive new information on the types of fluids that may have been involved in gold mineralization during the history of this important ore deposit. Samples from Tau Lekoa Mine, previously known as the Vaal Reefs No. 10 Shaft Mine in the Klerksdorp goldfield, and samples from the western part of Elandskraal Mine, previously known as the Elandsrand Gold Mine in the Carletonville gold-field, were analyzed. Three populations of predominantly secondary fluid inclusions are recognized. The earliest one comprises small, wispy-textured inclusions. The second involves a moderately saline fluid and the third one a CO 2 -rich fluid with variable CO 2 /H 2 O ratios. Microthermometry indicates fluid entrapment temperatures from 220 to 370 °C at 2 kbar. Tau Lekoa fluid inclusions are enriched in CO 2 compared to the CH 4 -enriched inclusions in Elandskraal samples. These differences are explained by comparatively more reducing conditions at the time of fluid entrapment at Elandskraal. Detailed textural studies indicate that fluid entrapment occurred at a late stage in the evolution of the Witwatersrand Basin. This is corroborated by previous argon chronological information as well as detailed mineralogical analysis of impact-related pseudotachylitic breccias, which also suggest that fluid migration occurred at the time of the Vredefort impact event. Thermodynamic calculations based on information gained from entrapped remnants of mineralizing fluids indicate that the solubility of gold in these fluids is, at 1–10 parts per billion, an order of magnitude lower than previously thought. These results cast doubt on the hydrothermal model for authigenic gold mineralization in the Witwatersrand Basin and favor a modified placer model.