Paleobiology, 26(4), 2000, pp. 578±605 Origination and extinction components of taxonomic diversity: Paleozoic and post-Paleozoic dynamics Mike Foote Abstract.ÐChanges in genus diversity within higher taxa of marine animals on the temporal scale of a few million years are more strongly correlated with changes in extinction rate than with chang- es in origination rate during the Paleozoic. After the Paleozoic the relative roles of origination and extinction in diversity dynamics are reversed. Metazoa as well as individual higher taxa shift from one mode of diversity dynamics to the other. The magnitude of taxonomic rates, the relative var- iance of origination and extinction rates, and the presence or absence of a long-term secular in- crease in diversity all fail to account for the shift in importance of origination and extinction in diversity changes. Origination and extinction rates both tend to be diversity-dependent, but dif- ferent modes of diversity-dependence may contribute to the change in diversity dynamics from the Paleozoic to the post-Paleozoic. During the Paleozoic, there is a weak tendency for extinction rates to be more diversity-dependent than origination rates, whereas after the Paleozoic the two rates are about equally diversity-dependent on average. Mike Foote. Department of the Geophysical Sciences, University of Chicago, Chicago, Illinois 60637. E-mail: [email protected] Accepted: 21 June 2000 Components of Diversity Change during of evolutionary innovations and ecological op- the Phanerozoic portunities (Walker and Valentine 1984; Gil- insky and Bambach 1987). In such a case, most How does variation in diversity during the of the variation in diversity within a clade history of life depend on variation in origi- would be attributable to variation in origina- nation and extinction rates? Any change in tion rate. Conversely, if variation in diversity taxonomic diversity necessarily implies a dif- is correlated primarily with variation in ex- ference between origination and extinction tinction rate, this might imply a characteristic rates, but by itself it says nothing about the rate of origination that is offset by external magnitude of the rates or how they may have perturbations leading to extinction. changed individually. Although diversity can Whether temporal variation in diversity change if the underlying taxonomic rates are more clearly re¯ects variation in origination constant, these rates in fact appear to have rate or variation in extinction rate is likely to ¯uctuated considerably over the scope of the be related to temporal and taxonomic scale. Phanerozoic (Raup and Sepkoski 1982; Van Some analyses of extant clades that are a few Valen 1984; Raup 1991; Foote 1994; Sepkoski million years old have suggested that recon- 1998). The central question I will address con- structed branching patterns are best explained cerns the relationship between these ¯uctua- by a model in which there is no extinction tions in origination and extinction rate and the (Hey 1992), although simulation studies have corresponding changes in diversity. Does di- shown that it may be dif®cult to identify past versity change tend to result more from extinction when only living representatives of change in origination rate or change in extinc- the clade are studied (Hey et al. 1998; cf. Raup tion rate, or is there no systematic pattern? The 1983; Nee et al. 1995; Harvey and Rambaut answer has obvious implications for diversity 1998). At the scale of animals as a whole over dynamics (Alroy 1998). For example, one the Phanerozoic, there has been a net diversity could imagine an extreme case in which ex- increase since the Paleozoic, implying an in- tinction rate within a clade is approximately crease in the difference between origination constant, perhaps for intrinsic biological rea- and extinction rates, but this increase has been sons, while origination rate varies as a result accompanied by a decline in both rates (Raup q 2000 The Paleontological Society. All rights reserved. 0094-8373/00/2604-0003/$1.00 DIVERSITY DYNAMICS 579 and Sepkoski 1982; Van Valen 1984). At the stratigraphic stages and series, I split the level of classes and orders over hundreds of Phanerozoic into 107 intervals in an attempt to millions of years, these higher taxa tend to ex- minimize the variance in estimated interval perience a secular decline in family origina- length; the mean interval duration is 5.1 m.y., tion rate, but extinction rate may be relatively with a standard deviation of 1.7 m.y. I also stationary (Van Valen 1985a; Van Valen and used an alternative subdivision of the Phan- Maiorana 1985; Gilinsky and Bambach 1987). erozoic (see Sepkoski 1996b; Foote and Sep- Thus, the long-term diversity history of many koski 1999) and found the results to be con- groups seems to be attributable to changes in sistent. The timescale is mainly that of Har- origination rate rather than extinction rate. land et al. (1990), with some modi®cations for There is considerable variation in rate super- the lower Paleozoic based on the timescales of imposed on these secular trends, however. It Tucker and McKerrow (1995) and Bowring is this ®ner-scale variation on which I will fo- and Erwin (1998). I will argue that inaccura- cus. Analysis of genera and families within cies in absolute dating are not responsible for higher taxa suggests that variation in extinc- the results presented here. All told, I included tion rate was more important than variation in 25,049 fossil genera whose ®rst and last oc- origination rate in determining diversity currences could both be fully resolved to one changes during the Paleozoic and that the rel- of the 107 intervals. Of these, 13,302 (53.1%) ative importance of extinction and origination have ranges that span more than one interval reversed from the Paleozoic to the Mesozoic. and thus could be used for analysis, as ex- Note that the question of temporal variation plained below. I included extant genera that which I address is distinct from that of vari- have a fossil record. I will show that excluding ation among biologic groups. A few studies them to remove potential biases resulting have attempted to factor differential diversi- from nearly complete knowledge of the living ®cation of related taxa into origination and ex- fauna has a relatively small effect on the re- tinction components. For example, Cheetham sults. In addition to analyzing the data as a and Jackson (1996) suggested that two coeval whole, I analyzed subsets corresponding to genera of cheilostome bryozoans have had the more diverse phyla, classes, and orders. similar histories of origination, and that the I measured diversity as the number of gen- difference in their respective species richness era known to cross interval boundaries (Ray- is therefore due to a difference in extinction mond and Metz 1995; Alroy 1996, 1998; Bam- rate. Conversely, Van Valen (1973) compared bach 1999). This provides an estimate of in- artiodactyl and perissodactyl mammals and stantaneous standing diversity that does not found similar extinction rates, suggesting that increase systematically with interval length the greater diversity of artiodactyls is attrib- and that, by ignoring single-interval genera, is utable to a higher origination rate. Another re- not expected to be as sensitive to variation in lated question concerns the relative impor- preservation rate as is the total number of gen- tance of stages of allopatric speciation in net era known from an interval (Pease 1985; Sep- rates of origination. Allmon (1992) has co- koski 1990, 1993, 1996b, 1997, 1998; Buzas and gently argued that differences in speciation Culver 1994, 1998; Raymond and Metz 1995; rate may be better understood mechanistically Harper 1996; Alroy 1996, 1998; Bambach 1999; if one investigates differences in rates of iso- Foote 2000). Diversity was measured on a log- late formation, isolate persistence, and isolate arithmic scale in order to monitor proportion- divergence. al change (Sepkoski 1991), but results are sim- ilar if an arithmetic scale is used. Data and Methods For each interval, I estimated extinction rate I analyzed stratigraphic range data from (q i) per lineage-million-years (Lmy) based on Sepkoski's unpublished compendium of the proportion of genera present at the start of Phanerozoic marine animal genera (see Sep- an interval that survived at least to the end of koski 1996b, 1997, 1998 for a description of the the interval. Similarly, I estimated origination data). By dividing and combining standard rate (p i) per Lmy based on the proportion of 580 MIKE FOOTE genera present at the end of an interval that in this study. Rate metrics that include single- had already originated by the start of the in- tons assign these taxa to both the origination terval. Thus, qÃi 52ln[(Nbt)/Nb]/Dt and pà i 5 and extinction tabulations for a given interval. 2ln[(Nbt)/Nt]/Dt, where Nb is the total num- If the number of singletons ¯uctuates arti®- ber of genera that cross the bottom interval cially as a result of varying quality of preser- boundary, Nt is the total number of genera that vation, then a spurious correlation between cross the top interval boundary, Nbt is the origination and extinction rates may result. number of genera that cross both boundaries, This can be shown theoretically (Foote 2000); and Dt is the duration of the interval (Foote it has also been shown empirically for Ceno- 1999: Appendix 7; Foote 2000). Although pà i zoic mammals (Alroy 1996, 1998). Such a spu- and qÃi are estimates of the true values pi and rious correlation could make it dif®cult to sep- qi, for simplicity I will hereinafter refer to arate diversity changes into their origination these estimates as pi and qi. These rate esti- and extinction components. Singletons would mates satisfy the requirements of branching not present such a problem if origination and 2qiiDt 2p Dt theory that Nbt 5 Nebt, Nbt 5 Ne , and extinction rates were equal within a given in- (pii2q )Dt Nt 5 Neb (Kendall 1948; Raup 1985).