We present a quantitative analysis of megafloral turnover across the Cretaceous/Paleogene boundary (K/T) based on the most complete record, which comes from the Williston Basin in southwestern North Dakota. More than 22,000 specimens of 353 species have been recovered from 161 localities in a stratigraphic section that is continuous across and temporally calibrated to the K/T and two paleomagnetic reversals. Floral composition changes dynamically during the Cretaceous, shifts sharply at the K/T, and is virtually static during the Paleocene. The K/T is associated with the loss of nearly all dominant species, a significant drop in species richness, and no subsequent recovery. Only 29 of 130 Cretaceous species that appear in more than one stratigraphic level (non-singletons) cross the K/T. Only 11 non-singletons appear first during the Paleocene. The survivors, most of which were minor elements of Cretaceous floras, dominate the impoverished Paleocene floras. Confidence intervals show that the range terminations of most Cretaceous plant taxa are well sampled. We infer that nearly all species with last appearances more than about 5 m below (approximately 70 Kyr before) the K/T truly disappeared before the boundary because of normal turnover dynamics and climate changes; these species should not be counted as K/T victims. Maxima of last appearances occur from 5 to 3 m below the K/T. Interpretation of these last appearances at a fine stratigraphic scale is problematic because of local facies changes, and megafloral data alone, even with confidence intervals, are not sufficient for precise location of an extinction horizon. For this purpose, we rely on high-resolution palynological data previously recovered from continuous facies in the same sections; these place a major plant extinction event precisely at the K/T impact horizon. Accordingly, we interpret the significant cluster of last appearances less than 5 m below the K/T as the signal of a real extinction at the K/T that is recorded slightly down section. A maximum estimate of plant extinction, based on species lost that were present in the uppermost 5 m of Cretaceous strata, is 57%. Palynological data, with higher stratigraphic but lower taxonomic resolution than the megafloral results, provide a minimum estimate of a 30% extinction. The 57% estimate is significantly lower than previous megafloral observations, but these were based on a larger thickness of latest Cretaceous strata, including most of a globally warm interval, and were less sensitive to turnover before the K/T. The loss of one-third to three-fifths of plant species supports a scenario of sudden ecosystem collapse, presumably caused by the Chicxulub impact.