Because more than one calibration is used, quartet dating (a molecular dating method) is thought to reduce error that might arise from a single calibration point. Within crocodylians, there is a strong correlation between calibration age and divergence estimate for five mitochondrial genes and one nuclear gene—estimates based on two Neogene calibrations are all younger than those based on two Paleogene calibrations, and estimates based on one Neogene and one Paleogene calibration are of intermediate age. Confidence limits on the youngest estimates exclude the oldest estimates, and in several cases they exclude known minimum divergences from fossil occurrences. Addition of time to the calibrations improves among-quartet and stratigraphic consistency, but not all kinds of modifications have the same impact; addition of uniform blocks of time to all calibrations efficiently increases among-quartet consistency, but with range extensions that more than double some of the Neogene calibrations. Modest increases in calibration age disproportionately impact divergence estimates based on later calibrations. Some among-quartet disparity might reflect calibration error, especially among caimans, but some range extensions necessary to improve consistency are unreasonably long. Quartet dating appears to systematically undercompensate branch length error with late calibrations and overcompensate it with early calibrations, but in all cases very reasonable results—alligatorid-crocodylid divergence in the Late Cretaceous and Alligator-caiman divergence at or near the Cretaceous-Tertiary boundary—are obtained when both a Neogene and a Paleogene calibration is used. This suggests that, given current likelihood models, the use of calibrations sampling different parts of a clade's history is the best strategy when using quartet dating.