107Syst. Biol. 48(1):107–118, 1999The F ossil Record of N orth American Mammals: Evidence for aP aleocene Evolutionary Radiatio nJOH NAL ROY1Department of Paleobiology, Smithsonian Institution, MRC 121, Washington, D.C. 20560, USAAb stract.— Paleontologists long have argued that the most important evolutionary radiation ofmammals occurred during the early Cenozoic, if not th at all eutherians originated from a singlecommon post-Cretaceous ancestor. Nonetheless, several recent molecular analyses claim to showth at because several interordinal splits occurred during the Cretaceous, a major therian radiationwas then underwa y. Th is claim conicts with statistical evidence from the well-sampled latestCretaceous and Cenozoic North American fossil record. Paleofaunal data conrm that there werefewer mammalian species during the latest Cretaceous than during any interval of the Cenozoic,and that a massive diversication took place during the early Paleocene, immediately after a massex tinction. Measurement data show that Cretaceous mammals were on average small and occupieda narrow range of body sizes; after the Cretaceous–Tertiary mass extinction, there was a rapid andperman ent shift in the mean. The fact that there was an early Cenozoic mammalian radiation isen tirely compatible with the existence of a few Cretaceous splits among mod ern mammal lineages.[B ody mass; Cenozoic; Cretaceous; diversication; extinction; Mammalia; molecular clock.]O ver the past few years, there has been anex plosion of interest in the early evolution-ary r adiation of mammals. Traditional sce-nar ios based mostly on paleontological datahave been challenged by inferences based onthe calibration of molecular phylogenies tonumer ical time (Hedges et al., 1996; Janke etal., 1997; Springer, 1997; Cooper and Fortey,1998; Kumar and Hedges, 1998). Some ofthese new molecular studies (e.g., Kumarand Hedges, 1998) purport not just to over-thr ow traditional higher-order phylogeneticgroupin gs, but also to show that a majord iversication of therian mammal s beganmuch earlier than previously thought, per-haps even in the Early Cretaceous.I f the fate of some other recent debatesin mammalian molecular systematics is aguid e, then some of the novel topologiesun dergirding th ese results may be incor-rec t. For example, Sullivan and Swofford(1997) h ave shown that a heated debateover the possible polyph yly of rodents (e.g.,D’Erchia et al., 1996) rested on inadequateanalys es. Regardless of topologies, most ofthe “Cretaceous radiation” research suffer sfrom using just one or two “clock” calibra-tion points (Hedges et al., 1996; Cooper andPenny, 1997; Janke et al., 1997; Kumar andH edges, 1998). Such c alibrations often yieldan omalous results that are defended by as-s ertions that all of the conicting, paleonto-log ically inferred dates of origin are simplytoo young.For example, Hedges et al. (1996: 227) jus-tify their decision to use a single Carbonifer-ous calibration point with a non sequitur—that there is a “long time span between theear liest [mammalian] fossils. . . and the rstappear ance of the modern orders.” In otherw ords, because the modern mammal orderss eem to Hedges et al. to have appeared longaf ter they diverged, the authors can justifyus ing a Carboniferous calibration point in-s tead, and because the Carboniferous pointis reasonable, they can infer that the modernmammal orders appeared long after they ac-tually diverged. Tautological reasoning likethis makes it impossible to move on to a rea-s oned discussion of the relative reliability ofmolecular clocks and paleontological data.For example, if the rst appearances of mod-ern mammal orders are even vaguely closeto their true divergence times, this stronglys uggests a speed up in the molecular clock1Present addres s and address for reprint requests:N ational Center for Ecological Analysis and Synthe-sis, University of California, 735 State Street, Suite300, Santa Barbara, California 93101, US A; E-mail:[email protected] YSTEMATIC BIOLOGY VOL .48of Hedges et al. (1996) sometime duringthe 245 m illion years (MY) between theC arboniferous and the Cretaceous–Ter tiary(K- T) bound ary.E ven the better studies have crucial aws.D espite having employed multiple calibra-tion points and trying tocorrect for variationin the clock speed, Springer (1997) arrived ata clock rate w ith alarmingly broad 95% con-dence limits of ± 13% (“XR adjusted”) or± 15% (“MRR adjusted”). Furthermore, allb ut one of the calibration points fell withinthe Cenozoic, forcing the interordinal diver-g ence times to be based largely on extrapo-lation instead of interpolation. This one Cre-taceous point was a supposed gure of 130m illion year s ago (MYA) for the marsupial–placen tal split. However, the original source(N ovacek, 1993) did not discuss the 130MYA date, which was read off of an ar-tis tically rendered text gure that clearlyimplied the absence of any concrete evi-d ence for a split before 98 MY. The youngerd at e has been conrmed by more recent re-s earch (C ifelli et al., 1997). Changing to a 98MYA estimate—a minimum gure like allof the other ones used by Springer (1997)—in creases the c lock rate, and therefore de-creas es all the estimated divergences, by12% .Des pite such concerns, my pur pose is notto challenge the inference th at the basals plits among many therian orders occurreds ometime during the Cretaceous. Ins tead,I will make three simple points. Fir st, them odern paleontological literature has neverimplied that all therian or even eutherian or-d er s diverged from one common ancestorafter the Cretaceous. Any claim to the con-tr ary is a misinterpretation that makes them olecular results seem more novel than theyreally are. Second, most molecular studiesh ave failed to dene the idea of “radiation”or “diversication” in a rigorous manner,lead ing to inferences from data that are notreally relevant to the debate. Finally, clearerd enitions imply that only two biologicalpatter ns are of interest in this discussion:ch anges through time in the overall num-ber of species, and changes through time inthe distribution of morphologies (or otherattr ibutes) across those species.The fossil record does provide clear-cutev idence regarding both of these patterns. Its how s that Cretaceous mammals were tax-on omically depauperate and morphologi-cally uniform, and that the most