A timescale is necessary for estimating rates of molecular and morphological change in organisms and for interpreting patterns of macroevolution and biogeography(1-9). Traditionally, these times have been obtained from the fossil record, where the earliest representatives of two lineages establish a minimum time of divergence of these lineages(10). The clock-like accumulation of sequence differences in some genes provides an alternative method(11) by which the mean divergence time can be estimated. Estimates from single genes may have large statistical errors, but multiple genes can be studied to obtain a more reliable estimate of divergence time(1,12-13). However, until recently, the number of genes available for estimation of divergence time has been limited. Here we present divergence-time estimates for mammalian orders and major lineages of vertebrates, from an analysis of 658 nuclear genes. The molecular times agree with most early (Palaeozoic) and late (Cenozoic) fossil-based times, but indicate major gaps in the Mesozoic fossil record. At least five lineages of placental mammals arose more than 100 million years ago, and most of the modern orders seem to have diversified before the Cretaceous/Tertiary extinction of the dinosaurs.