Fossils are widely used as calibration points in molecular-based dating studies, but their placement on a phylogenetic tree of extant species is always highly problematic. We explore some of the problems linked to calibration with fossils, in particular their position on the tree, and emphasize the use of multiple calibration points to obtain better estimates. We use a phylogenetic analysis of Betulaceae based on nuclear ribosomal DNA sequences (5S spacer and ITS) as a case study and estimated divergence times within the family using the nonparametric rate smoothing method of Sanderson and five calibration points from the extensive fossil record of this family. To assess the effects of assumptions relating to the positions of key fossils with respect to stem lineages versus crown groups, we calculated age estimates by placing each fossil subsequently on the stem lineage node and crown group node, and then determined the median value of the resulting ten estimates for each node. Using maximum likelihood and DELTRAN and ACCTRAN parsimony branch lengths, we found that the age of the crown group and stem lineage of Betulaceae vary from 115.2 to 130.6 million years and 211.2 to 302.6 million years (Aptian or before). These results are older than current paleobotanical data. We calculated paleobotanical confidence intervals using methods based on the occurrence of fossils on a stratigraphic column and the lengths of the gaps between these occurrences. We apply these methods to the fossil record of Alnus and related extinct genera; however, only in some cases were molecular- and fossil-based age estimates reconciled.