q 2004 The Paleontological Society. All rights reserved. 0094-8373/04/3004-0003/$1.00

Paleobiology, 30(4), 2004, pp. 522–542

Origination, extinction, and mass depletions of marine diversity

Richard K. Bambach, Andrew H. Knoll, and Steve C. Wang

Abstract.—In post-Cambrian time, five events—the end-Ordovician, end-Frasnian in the Late De-vonian, end-Permian, end-Triassic, and end-Cretaceous—are commonly grouped as the ‘‘big five’’global intervals of mass extinction. Plotted by magnitude, extinction intensities for all Phanerozoicsubstages show a continuous distribution, with the five traditionally recognized mass extinctionslocated in the upper tail. Plotted by time, however, proportional extinctions clearly divide the Phan-erozoic Eon into six stratigraphically coherent intervals of alternating high and low extinction in-tensity. These stratigraphic neighborhoods provide a temporal context for evaluating the intensityof extinction during the ‘‘big five’’ events. Compared with other stages and substages in the sameneighborhood, only the end-Ordovician, end-Permian, and end-Cretaceous extinction intensitiesappear as outliers. Moreover, when origination and extinction are considered together, only thesethree of the ‘‘big five’’ events appear to have been generated exclusively by elevated extinction. Loworigination contributed more than high extinction to the marked loss of diversity in the late Fras-nian and at the end of the Triassic. Therefore, whereas the ‘‘big five’’ events are clearly times whendiversity suffered mass depletion, only those at the end of the Ordovician, Permian, and Cretaceousperiods unequivocally qualify as globally distinct mass extinctions. Each of the three has a uniquepattern of extinction, and the diversity dynamics of these events differ, as well, from the other twomajor diversity depletions. As mass depletions of diversity have no common effect, common cau-sation seems unlikely.

Richard K. Bambach and Andrew H. Knoll. Botanical Museum, Harvard University, 26 Oxford Street,Cambridge, Massachusetts 02138. E-mail: rbambach@oeb.harvard.edu

Steve C. Wang. Department of Mathematics and Statistics, Swarthmore College, 500 College Avenue,Swarthmore, Pennsylvania 19081. E-mail: scwang@swarthmore.edu

Accepted: 15 February 2004

Introduction:Mass versus ‘‘Background’’ Extinction

The idea that mass extinctions stand out asa class of events separate from the range of‘‘normal’’ or ‘‘background’’ extinctions thatcharacterize most of the geological recordoriginated with the work of Norman Newell(1962, 1963, 1967) and crystallized throughthe quantitative analysis by Raup and Sepko-ski (1982). Using Sepkoski’s compilation ofstratigraphic ranges temporally resolved tothe stage level for marine families of all ani-mal taxa (Sepkoski 1982), Raup and Sepkoskiidentified five mass extinctions: the end-Or-dovician (Ashgillian); Late Devonian (includ-ing the Frasnian/Famennian boundary); end-Permian (Guadalupian and Djhulfian togeth-er); end-Triassic (Late Norian or Rhaetian);and end-Cretaceous (Maastrichtian). Thesehave become known informally as ‘‘the big

* This paper is dedicated to the memory of Stephen JayGould (1941–2002).

five’’ mass extinctions. Each had been notedearlier by Newell (1967).

Raup and Sepkoski plotted the extinctionrate (number of families going extinct per mil-lion years) for each stage in time order and as-sumed that apparent outliers on the arithme-tic plot were true statistical outliers (and notaberrations produced by errors in the time-scale), thus identifying a distinct class of massextinctions. They also calculated a 95% confi-dence interval around the remaining ‘‘back-ground’’ extinction stages and demonstrateda secular decrease in extinction rates overtime. Quinn (1983), however, showed that theextinction data were highly skewed and cor-rectly pointed out that the assumption of anormal distribution for ‘‘background’’ extinc-tion data was not valid. When log-trans-formed, the distribution of extinction rateswas approximately normal and did not have asuite of outliers that could be categorized un-ambiguously as a class of mass extinctionsseparate from the overall ‘‘background’’ dis-tribution. Bambach and Gilinsky (1986) dem-

523ORIGINATION, EXTINCTION, AND DIVERSITY DEPELETIONS

onstrated that this was also the case for severalother extinction metrics, including propor-tional metrics by interval that are not subjectto time-scale error. For all metrics the distri-bution of extinction intensity grades smoothlyfrom lowest to highest values with no discretebreak between ‘‘background’’ and presumedmass extinction intervals. Indeed, Raup (1991)developed his view of the Phanerozoic killcurve on the basis of this continuity of extinc-tion magnitudes. Bambach and Gilinsky(1986) did, however, support the finding thatextinction intensities (and origination inten-sities, as well) declined during the Phanero-zoic, a conclusion discussed more fully by Gil-insky (1994).

Despite the evidence that intensities of ex-tinction form a continuum, the five eventsspecified by Raup and Sepkoski (1982) contin-ue to be labeled ‘‘mass extinctions’’ (Finney etal. 1999; McGhee 2001; Wignall and Twitchett1996; Palfy et al. 2000; MacLeod and Keller1996). Current threats to biodiversity haveeven been labeled ‘‘the sixth extinction’’ (Lea-key and Lewin 1995). It may be that magni-tude alone is sufficient to justify the term‘‘mass extinction’’ because events with suchpronounced loss of diversity are rare, withwaiting times of about 100 million years (Raup1991). But is there any reason to think of theseevents as a separate class of events, rather thanas the uncommon upper tail of a continuousdistribution?

Wang (2003) recently identified three con-cepts that must be considered individuallywhen we ask whether putative mass extinc-tions grade continuously into the range ofbackground extinction: continuity of cause,continuity of effect, and continuity of magni-tude. Continuity of cause would be demon-strated if candidate mass extinctions could beshown to be driven by the same processes thatare responsible for background extinction, al-beit operating at increased intensity or overlarger areas. Continuity of effect would be es-tablished if background and mass extinctionsexhibited common patterns of selectivity ontaxonomic, functional, ecological, or othergrounds. And continuity of magnitude wouldexist if the distribution of intensities of massextinctions graded smoothly and continuous-

ly into intensities of background extinction.Although continuity of magnitude appears tohave been demonstrated by Quinn (1983) andBambach and Gilinsky (1986), and was as-sumed by Raup (1991) in his kill curve anal-ysis for the Phanerozoic, several factors led usto reconsider the status of these intervals.These include (a) the rarity of the largest ex-tinction intensities, (b) the possibility that theydo not share continuity of cause or effect withall other intervals, and (c) the fact that origi-nation and extinction have rarely been consid-ered together in examining patterns of diver-sity change.

In the following sections, we reevaluate thecontinuity or discontinuity of magnitude andthen briefly consider whether events thatmight be regarded as mass extinctions can beunified by effect or cause. We test whether anyof the ‘‘big five’’ events are differentiable fromthe distribution of other extinction intensities,explore the role of origination as well as ex-tinction in diversity changes associated withthese five intervals, and comment on the dif-ferences, as well as similarities, among the in-tervals. The upshot will be that although thefive intervals in question—the end-Ordovi-cian, Late Devonian, end-Permian, end-Trias-sic, and end-Cretaceous—are the five intervalswith the greatest diversity loss in the Phan-erozoic, they share little else in common. Onlythree were driven predominantly by extinc-tion and even they display distinctly differentpatterns of diversity change, implying thatthese events are not related by continuity ofeffect or cause.

Tracking Genus Diversity

Figure 1 illustrates the history of marine ge-nus diversity through Phanerozoic time. Thedata were compiled by using a computerizedsorting routine written by Jack Sepkoski andmodified by J. Bret Bennington to tabulate ge-nus diversity for 107 stages and substages us-ing Jack Sepkoski’s unpublished tabulation ofthe stratigraphic ranges of genera as of 1996.Although the Paleontological Research Insti-tution has recently published the raw genusranges from a later version (1998) of Sepko-ski’s compilation (Sepkoski 2002), this publi-

524 RICHARD K. BAMBACH ET AL.

FIGURE 1. Diversity and diversity turnover of marine genera by interval through the Phanerozoic. The five majorpost-Cambrian diversity depletions are highlighted. The heavy line connects the data on number of genera crossingeach interval boundary. The line directly connecting the numbers of boundary-crossing genera follows the path ofminimum likely standing diversity, regarded as the minimum diversity because origination and extinction wouldhave to work in exact lock-step to follow that diversity path. The peaked dotted line represents genus turnoverwithin each interval. The rising part of each peak represents all genus originations (first occurrences) reported fromthe interval. The peak records the total number of genera reported from the time interval. The descending part ofthe peak represents the number of extinctions (last records) of genera in the interval. The magnitudes of the peakscompared with the minimum standing diversity at interval boundaries represent the degree of faunal turnover inthe intervals.

cation simply lists all the genera and does notnumerically tabulate the data on diversity.

Genus diversity follows the general patternestablished in the ‘‘consensus paper’’ of Sep-koski et al. (1981) and best known from Sep-koski’s (1981: Fig. 5) widely reproduced fam-ily diversity curve. In the Paleozoic we see the‘‘Cambrian Explosion’’ (the increase in diver-sity in the Early Cambrian), a Middle and LateCambrian ‘‘plateau’’ of diversity, and the Or-dovician Radiation, followed by the long in-terval of fluctuating, but non-trending, diver-sity that began in the Caradocian and lastedfor the rest of the era. Diversity changes dur-ing this ‘‘Paleozoic Plateau’’ include three ofthe ‘‘big five’’ diversity depletions, the end-Ordovician, the Late Devonian, and the end-Permian events. The post-Paleozoic is charac-terized by nearly continuous diversity in-crease, interrupted by the other two ‘‘big five’’diversity depletions, the end-Triassic and thesharp, era-bounding end-Cretaceous events.Note that because we tabulated subgenera ofmollusks, resolved the data to the substagelevel, and emphasized the diversity at intervalboundaries rather than the total diversity

within each interval, the apparent Cenozoicincrease in diversity is proportionately greaterthan that illustrated by Sepkoski for families.

There are some concerns that the apparent-ly large Cenozoic increase in diversity may bepartly artifactual. However, recent analyses(Bush and Bambach in press on alpha diver-sity; Jablonski et al. 2003 on Pull of the Recent;and Bush et al. 2001, in press on techniques ofsample-standardization) demonstrate that anincrease in Cenozoic diversity is strongly in-dicated, although the exact amount is still un-clear (Jackson and Johnson 2001). The ‘‘con-sensus paper’’ (Sepkoski et al. 1981) was puttogether because several of the data sets avoid-ed or mitigated some of the biases, such as im-perfections of the geologic record, that were ofconcern then and still are (Peters and Foote2001). Although every potential problemshould be analyzed and improvement in thedata is necessary, it still appears, as was con-cluded then, that the diversity signal is stron-ger than the noise.

Figure 1 accounts for all the data in the Sep-koski genus compilation (see caption for fullexplanation). Many diversity curves use total

525ORIGINATION, EXTINCTION, AND DIVERSITY DEPELETIONS

diversity as the recorded data (this is true forpublished illustrations of Sepkoski’s familycurve, for example). If one mentally ‘‘connectsthe dots’’ of total diversity peaks in Figure 1,it is clear that the general shape of a curve con-necting total diversities would be very similarto the boundary-crossing diversity empha-sized here. Mid-Devonian diversity would ap-pear higher than Late Ordovician diversity,Carboniferous diversity would fluctuate more,and the upturn in diversity in the mid-Cre-taceous would be sharper than shown by thestanding diversity plot. The general pattern,however, would be the same.

We use the boundary-crossing standing di-versity as the preferred representation of di-versity and diversity change through time. Weknow that boundary-crossing diversity is notanomalous compared with total diversity be-cause the trend of boundary-crossing diversi-ty follows the general path of total diversity.Two factors make us prefer it to total diver-sity. First, it is the only measure we have ofactual standing diversity. Total diversity in aninterval was not the actual standing diversityat any time because it is unlikely that all orig-inations occurred in an interval before any ex-tinction. However, the recorded diversity atthe boundaries of each interval, calculated bysubtracting all extinctions in the interval fromthe total diversity in the interval, is a directmeasure of standing diversity at intervalboundaries. The other compelling reason isthat change in diversity is shown best by com-paring diversity at the start of different inter-vals.

Diversity is a function of both originationand extinction. The peaks of turnover withineach interval in Figure 1 reveal how much var-iation of diversity can occur in any interval,but comparing standing diversities at intervalboundaries tells us whether origination andextinction are in balance (little or no change ofdiversity from one boundary to the next) orwhether either origination or extinction dom-inated during an interval (origination domi-nating if boundary-crossing diversity increas-es, extinction being more important if bound-ary-crossing diversity decreases). In plots oftotal diversity, a predominance of extinctionover origination in one interval could be

masked by an increase in origination in thenext. Total diversity might appear unchangedbetween the two intervals because originationin a rapid recovery from an extinction eventcould make total diversity in the succeedinginterval equal to that in the previous one, con-cealing the low diversity at the start of the in-terval. For example, although extinction ex-ceeded origination in each of the last two in-tervals of the Silurian and diversity appears tohave been lost between the Ludlovian and Pri-dolian when looking at total diversity, the lowpoint of diversity at the end of the Silurian(end-Pridolian) is masked in the total diver-sity curve because, in the Gedinnian, the firstinterval of the Devonian, origination was high,causing total diversity to exceed that of thePridolian. Boundary-crossing diversity notonly approximates standing diversity, but italso gives a clearer representation of the con-sequences of within-interval evolutionary dy-namics than that provided by summed totaldiversity data for whole intervals.

The ‘‘Big Five’’ as Diversity Depletions

Inspection of Figure 1 reveals that, althoughdiversity decreased slightly on several otheroccasions, there are only five post-Cambrianintervals when diversity decreased markedly:(1) at the end of the Ordovician, (2) during theMiddle and Late Devonian, (3) during the LatePermian, (4) at the end of the Triassic, and (5)at the end of the Cretaceous. The end-Triassicdecrease does not look as large as the otherfour, but standing diversity throughout theTriassic was lower than at any other time afterthe mid-Ordovician, so the proportional de-crease in the latest Triassic is quite compara-ble to the other four major diversity deple-tions. Two Cambrian intervals are also notedon Figure 1 (the late Botomian [LB] and earlylate Middle Cambrian [ELM]). These weretimes of relatively small numerical change indiversity but high proportional diversity loss.Because the evolutionary dynamics of theCambrian and Early Ordovician are unusualwe will consider them separately; the bulk ofthis paper emphasizes the post-Arenig por-tion of the Phanerozoic.

Because it is hard to judge the proportionalmagnitude of diversity change from a plot of

526 RICHARD K. BAMBACH ET AL.

FIGURE 2. Proportion of gain or loss of genus diversity from the Caradoc to the Plio-Pleistocene. The five majordiversity depletions (decrease greater than 20%) are numbered. Symmetrical lines are drawn at 213.5% and 113.5%(based on the sixth largest diversity decrease) to indicate the range that might be regarded as ‘‘background’’ fluc-tuation in diversity. Intervals with greater than 13.5% diversity increase are common only after major diversitydepletions.

absolute values, a plot of proportion of gain orloss of diversity, rather than a plot of the num-bers of genera as such, is desirable. A displayof numbers of taxa, as in Figure 1, is useful forillustrating the pattern of change in diversity,but the numbers represent only those taxa dis-covered in the fossil record and are not a com-plete record of all the taxa that existed. Pro-portional diversity change is what we seek tounderstand here—how episodes of diversityloss affected the whole biota. The question, ineffect, concerns the importance of an event inthe context of its time, not just how many taxawere involved.

Figure 2 shows the proportional gain or lossin the number of genera during each stage orsubstage interval, starting with the Early Car-adocian in the Middle Ordovician after thelarge proportional increases in diversity as-sociated with the Cambrian Explosion and theOrdovician Radiation were over. The changein diversity is calculated as a proportionalchange by subtracting the number of genera atthe start of each interval (the standing diver-sity at the boundary between the interval andits preceding interval) from the number ofgenera at the end of the interval (the standingdiversity at the boundary between the intervaland its succeeding interval) and dividing that

number (the change in the number of generafrom the start to the end of the interval) by thenumber of genera at the start of the interval.For example, if 500 genera pass from intervalone to interval two and 600 genera pass frominterval two to interval three, then there were100 more originations than extinctions duringinterval two, with a gain in diversity of 100genera, a proportional increase of 10.200.Likewise, a decrease from 600 to 500 generaduring an interval (a loss of 100 genera as aresult of 100 more extinctions than origina-tions) would be a proportional decrease of20.167. As noted above, an advantage of de-termining standing diversity at intervalboundaries is that one can follow the balanceof origination and extinction as it influencesdiversity change, something not possiblewhen only tabulating total diversity for eachinterval.

Figure 2 shows that the five intervals al-ready well known as the classic ‘‘big five’’mass extinctions (with the Guadalupian aswell as the end-Permian Djhulfian included inthe major later Permian diversity decrease[Stanley and Yang 1994]) are the only post-Llandeilian intervals with more than a 20%proportional loss of genus diversity. In fact,there is a gap of 8% between the largest loss

527ORIGINATION, EXTINCTION, AND DIVERSITY DEPELETIONS

of diversity not included in the ‘‘big five’’ andthe smallest loss within the ‘‘big five,’’ but nogaps of as much as 2% occur between anysmaller values when arranged in rank order.These five intervals were certainly times ofmass depletion of diversity, but can we justifyregarding any of them as times of mass ex-tinction different in cause, effect, or magni-tude from ‘‘background’’ extinction?

Large-Magnitude Proportional Increasesin Diversity

As a side issue, but one related to propor-tional diversity change and its timing, it is in-teresting to note that the only times when pro-portional increase of diversity exceeds 13.5%are during the Cambrian Explosion, the Or-dovician Radiation (just ending in the earlyCaradocian at the start of Fig. 2), in the im-mediate aftermath of each of the five ‘‘massdepletions’’ of diversity, and briefly (single in-tervals only) in the Late Cretaceous (Turoni-an) and Neogene (early Miocene) (Fig. 2). Al-though origination rates are not unusuallyhigh in these intervals (no outliers for origi-nation are found in an analysis of originationproportions at these times), the combinationof higher origination and lower extinctionduring the ‘‘recovery’’ phase after diversitydepletion does mark these intervals as timesof unusually great proportional increase in di-versity. These are not necessarily times ofbroad transgression or otherwise better rep-resentation of the marine record, so higherorigination rates in the wake of major diver-sity depletions may reflect recovery from un-usually low diversity and not just the effect ofimproved record availability, a possibilityraised by Peters and Foote (2001).

Testing for Continuity of Magnitudeof Extinction

We tested for continuity or discontinuity ofmagnitude of extinction (i.e., whether or notthe distribution of intensities of apparentmass extinctions grade smoothly and contin-uously into intensities of background extinc-tion) in two ways. (1) We tested whether thereis a smooth continuous distribution of mag-nitudes of extinction intensity with no strongvariation in the upper tail of the distribution,

first for the whole Phanerozoic and second forthe time after the after the ‘‘Cambrian Pla-teau’’ of low diversity and high turnover. (2)We compared magnitudes of extinction foreach interval against the distribution of mag-nitudes of extinction in the particular segmentof the timescale, based on average high or lowextinction rates, to which the interval belongs.Only if an interval satisfies both criteria, thatis, if the interval is not part of a continuoussmooth distribution of magnitudes, and if theinterval also appears as an ‘‘outlier’’ in mag-nitude compared with the other intervals inits particular segment of the timescale, do weregard it as a ‘‘true’’ global mass extinction,different in magnitude from the bulk of as-sociated stratigraphic intervals.

How Continuous Are the Values of ExtinctionIntensity? As noted above, several analyseshave concluded that the distribution of extinc-tion intensities is apparently continuous(Quinn 1983, Bambach and Gilinsky 1986,Raup 1991, Wang 2003). Figure 3A shows thiseffect for proportion of genus extinction ar-ranged in rank order for the 107 stages andsubstages of the Phanerozoic as tabulated inour version of Sepkoski’s genus database.

Cambrian and Early Ordovician extinctionproportions, however, were consistently high(Table 1, Fig. 4). Sixteen of 19 Cambrian andEarly Ordovician intervals have extinction in-tensities that fall within the top quartile of allPhanerozoic intervals (Fig. 3A). Originationwas unusually high, as well, during this in-terval. Thus, whereas taxonomic turnover atthe genus level was great, overall diversity didnot fluctuate wildly (Fig. 1 and discussion be-low). These turnover rates are not like those ofmuch of the later Phanerozoic. For example,the decrease in proportion of extinction ob-served in the Early Ordovician (Fig. 4A) wasnot produced by the radiation of taxa withlower extinction rates diluting continuinghigh extinction rates in the trilobites, whichdominated Cambrian diversity. Instead, ex-tinction proportions for trilobites, which hadconsistently exceeded those of the non-trilo-bite fauna from the origin of the clade in theearly Atdabanian though the early Arenigian(Foote 1988), dropped to the same level asnon-trilobite proportions of extinction during

528 RICHARD K. BAMBACH ET AL.

FIGURE 3. Proportions of genus extinction arranged in rank order by magnitude. A, All 107 intervals of the Phan-erozoic. Magnitudes from the Cambrian and Early Ordovician are highlighted. B, Middle Ordovician to Plio-Pleis-tocene values only. Higher magnitude intervals are labeled.

the Late Arenigian and remained at compa-rable levels thereafter (Fig. 4C). Clearly some-thing changed for trilobite evolutionary dy-namics in the later part of the Early Ordovi-cian. Although the change was not as dramaticfor the non-trilobite fauna, extinction propor-tions for that fraction of the fauna, which hadbeen over 30% in two-thirds of the intervals ofthe Cambrian and Early Ordovician, droppedto levels below 30% and remained low untilthe Middle Silurian, except for the end-Or-dovician late Ashgillian extinction event. Pro-portions of extinction never returned consis-tently to the high levels common in the Cam-

brian and Early Ordovician, even when pro-portions of extinction increased for extendedintervals, such as from the Middle Silurian tothe mid-Carboniferous—this held for trilo-bites and non-trilobites alike.

We do not yet understand why turnoverrates should have been so high during theCambrian and Early Ordovician. Perhaps ear-ly animals were more vulnerable to extinctionfor functional reasons—many Cambrian ani-mals belonged to stem rather than crowngroups of bilaterian phyla and classes (Buddand Jensen 2000). Perhaps the low diversity ofCambrian and Early Ordovician animals con-

529ORIGINATION, EXTINCTION, AND DIVERSITY DEPELETIONS

TA

BL

E1.

Dat

aon

pro

por

tion

sof

orig

inat

ion

and

exti

nct

ion

for

maj

orin

terv

als

ofth

eP

han

eroz

oic

bas

edon

gen

eral

mag

nit

ud

eof

exti

nct

ion

.

Maj

orin

terv

al

Nu

mb

erof

con

tain

edin

terv

als

Min

imu

mp

rop

orti

onof

orig

inat

ion

Mea

np

rop

or-

tion

ofor

igin

atio

n*

Max

imu

mp

rop

orti

onof

orig

inat

ion

Min

imu

mp

rop

orti

onof

exti

nct

ion

Mea

np

rop

orti

onof

exti

nct

ion

*

Max

imu

mp

rop

orti

onof

exti

nct

ion

Om

itte

dp

rop

orti

ons

ofex

tin

ctio

n*

Cam

bria

n&

Ear

lyO

rdov

icia

nM

idd

leO

rdov

icia

n–E

arly

Silu

rian

Mid

Silu

rian

–Ear

lyC

arb

onif

erou

s

19 11 18

0.26

80.

133

0.15

5

0.53

90.

245

0.28

6

0.69

50.

406

0.44

1

0.27

00.

114

0.19

1

0.46

00.

174

0.28

7

0.64

40.

273

0.38

9

—0.

569

0.34

7L

ate

Car

bon

ifer

ous

and

Per

mia

nT

rias

sic

Jura

ssic

thro

ugh

Ple

isto

cen

e

14 6 39

0.04

40.

211

0.08

5

0.15

70.

413

0.18

5

0.31

50.

563

0.32

8

0.05

40.

281

0.04

3

0.12

80.

341

0.12

3

0.22

20.

452

0.25

3

0.55

60.

686

0.46

80.

474

*T

he

mea

ns

for

pro

por

tion

ofor

igin

atio

nar

eca

lcu

late

dby

usi

ng

the

pro

por

tion

sof

orig

inat

ion

from

all

the

stag

ean

dsu

bst

age

inte

rval

sin

each

maj

or

inte

rval

.Th

em

ean

sfo

rp

rop

orti

onof

exti

nct

ion

are

calc

ula

ted

byom

itti

ng

the

stag

eor

sub

stag

ein

terv

als

asso

ciat

edw

ith

the

‘‘bi

gfi

ve’’

div

ersi

tyd

eple

tion

s.T

he

omit

ted

pro

por

tion

sof

exti

nct

ion

are

list

edin

the

colu

mn

atth

efa

rri

ght.

Th

eyar

eth

ose

for

the

late

Ash

gil

lian

(Ord

ovic

ian

),la

teF

rasn

ian

(Dev

onia

n),

Gu

adal

up

ian

and

Djh

ulfi

an(P

erm

ian

),la

teN

oria

np

lus

Rh

aeti

an(T

rias

sic)

,an

dM

aast

rich

tian

(Cre

tace

ous)

.

FIGURE 4. Genus extinction by interval in stratigraphic(time) order. A, Proportion of genus extinction. TheCambrian and Early Ordovician concentration of veryhigh proportions is highlighted. That concentration isfurther emphasized because only six of the 88 post-Ar-enigian intervals rise above the dashed horizontal lineat 40% genus extinction. B, Natural logarithms of theproportion of genus extinction. Intervals of generallyhigh or low proportion of extinction are emphasized bythe fluctuation of the majority of points within blocks oftime intervals above and below the horizontal linedrawn at the lowest proportion in the Cambrian andLate Ordovician. Those large-scale long-term fluctua-tions are, of course, also visible in the plot of proportions(Fig. 4A). The dashed line is as in Figure 4A (the naturallog of 40% genus extinction). Intervals identified in Fig-ure 5 as forming a group of outliers (the late Ashgillianin the Ordovician, the Guadalupian and Djhulfian at theend of the Permian, and the Maastrichtian at the end ofthe Cretaceous) are highlighted. C, Proportion of genusextinction of trilobites and non-trilobites during theCambrian and Ordovician (except the end-Ordovician).Note the shift of trilobites to magnitudes similar to non-trilobites during the last two intervals of the Early Or-dovician and the lower proportions of all after that.

530 RICHARD K. BAMBACH ET AL.

tributed to low ecosystem stability (Bambachet al. 2002); many modes of life now heavilyoccupied were either vacant or contained fewtaxa (Bambach 1983, 1985). Or perhaps phys-ical environments experienced unusual per-turbation during this interval, as might be in-ferred from the carbon isotope record (Brasierand Sukhov 1998; Saltzman et al. 2000).

Whether we know the cause or not, the ob-served patterns of turnover show that Cam-brian and Early Ordovician extinction ratesform a distinct stratigraphic grouping that dif-fers from the remainder of the PhanerozoicEon. Statistically, it is extremely unlikely thatthe proportion of extinction in as many as 16of the 19 Cambrian and Early Ordovician in-tervals would fall within the top quartile of allPhanerozoic values if there were nothing un-usual about that segment of time (p 50.000000002, calculated by using the hyper-geometric distribution). Thus, we removedthese data and replotted the distribution of ex-tinction intensities.

The resulting rank order distribution ofmid-Ordovician to Recent extinction intensi-ties (Fig. 3B) retains the original smoothlycontinuous appearance except for six intervalsat the upper end: four of the ‘‘big five’’ (theend-Ordovician, end-Permian, end-Triassic,and end-Cretaceous) and the two intervalsthat bracket the terminal Permian interval.The high Guadalupian intensity just beforethe end-Permian Djhulfian likely combinestrue extinction (Stanley and Yang 1994; Jin etal. 2000) with Signor-Lipps range truncations(Signor and Lipps 1982; Jablonski 1986; Raup1987) from the exceptionally severe end-Perm-ian event. The initial Triassic interval (the In-duan) was a time of low diversity followingthe end-Permian devastation, and the cause ofits high extinction intensity has yet to be de-termined.

The distribution of extinction intensities forthe post-Arenig (Fig. 3B) still can be regardedas a continuous distribution, but one that ismore skewed than that for the entire Phaner-ozoic—the coefficient of skewness for the totalPhanerozoic was 0.89 whereas that for thepost-Arenig is 1.37. However, it is extremelyunlikely that removing 19 randomly selectedintervals (consecutive or non-consecutive)

would increase the coefficient of skewness tosuch an extent (p , 0.0001, calculated by sim-ulation). The sparseness of the remaining highvalues raises the question of whether thosemagnitudes are actually outliers that can beregarded as a separate group from the bulk oflower values.

Do All Intervals That Appear Discontinuouswith the Bulk of Post-Arenig Intervals Form aGroup of Outliers When Compared with Their ‘‘Lo-cal’’ Segment of Geologic Time? We can test di-rectly for the continuity of magnitude be-tween background and potential mass extinc-tions in a temporal context. Because the focusof this paper is on the ‘‘big five’’ mass deple-tions, which are all post-Arenig in age and be-cause the Cambrian and Early Ordovician arenot comparable in evolutionary dynamics (forwhatever reason, as noted above), we restrict-ed the following analysis to the post-Arenigportion of the Phanerozoic.

We tested for continuity of magnitude be-tween background and mass extinctions overthe 88 post-Arenig intervals. To control for thefluctuation in local extinction regimes appar-ent in the grouping of intervals in Table 1 (forwhich the boundaries were chosen where thechange in extinction proportions seen in Fig.4 were greatest) we used time-adjusted datarather than the raw extinction intensities. Todo this, we first fit a smooth lowess curve tothe raw extinction intensities (Fig. 5A). Low-ess (‘‘Locally Weighted Scatterplot Smoother’’[Cleveland 1979]) is a nonlinear regressionmethod that smoothes out the noise in a timeseries or scatterplot in order to emphasize thesignal. To use lowess, one must first choose thevalue of a ‘‘bandwidth’’ parameter controllingthe degree of smoothing. We used a band-width of 11%, meaning that the smoothed val-ue for each interval is calculated as a weightedaverage of 11% of the surrounding intervals,with the intervals closest in time weightedmore heavily. We chose this value because theaverage length of a geologic time period in the470 million years after the Arenig is 52 millionyears, and 11% of 470 million is 52 million.Thus the smoothed value for each intervaltakes into account the extinction intensities ofthe surrounding intervals for a time equiva-lent to a geological period.

531ORIGINATION, EXTINCTION, AND DIVERSITY DEPELETIONS

FIGURE 5. Statistical analysis of post-Arenig extinctionmagnitudes. A, Nonlinear lowess regression with 11%(52-million-year) bandwidth, revealing long-term fluc-tuation in extinction intensities. B, Histogram of resid-uals from nonlinear lowess regression with intervalshaving intensities comprising a statistically significantsecond mode (p 5 0.028) identified. The Induan and theupper Norian are the last two intervals at high end tailof the main mode (time-adjusted values of 0.168 and0.204), but they still lie well below the range of the sec-ond mode (0.307–0.416). The late Frasnian time-adjust-ed value of 0.062 is only the twelfth highest in the mainmode and is one of ten values in the second bin abovethe mode.

We then calculated the residuals, the differ-ence between each interval’s extinction inten-sity and its lowess smoothed value. Usingthese residuals, we applied the critical band-width test (Silverman 1981), previously usedby Wang (2003) to test for continuity of mag-nitude over the entire Phanerozoic. The ideaof this test, in this context, is as follows: Sup-pose there is a curve (i.e., a probability densityfunction) that models an underlying processfrom which extinction intensities are gener-ated. If this extinction intensity curve wereunimodal, we would infer that mass extinc-tions are the right tail of a continuum of ex-tinction intensities—that background andmass extinctions are continuous in magnitude.

On the other hand, a bimodal extinction in-tensity curve would suggest that backgroundand mass extinctions are discontinuous inmagnitude. The critical bandwidth test esti-mates the shape of the extinction intensitycurve and tests whether there is significant ev-idence to reject the null hypothesis that thecurve is unimodal. We found statistically sig-nificant evidence (p 5 0.028) for the existenceof a second mode in the right tail of the dis-tribution of extinction intensities (Fig. 5B). Thefour intervals that make up this second modeare, from highest to lowest, the Djhulfian,Maastrichtian, Upper Ashgillian, and Guad-alupian, confirming that only the end-Ordo-vician, end-Permian, and end-Cretaceoushave extinction magnitudes that would not beexpected in their local context and are not partof a continuum of extinction magnitudes.

Pattern of Temporal Change in Proportion of Ge-nus Extinction. Figure 4 plots the proportionof extinction in each interval arranged in tem-poral order. Although the numerous high val-ues of extinction intensity in the early and midPaleozoic and the consistently lower intensi-ties in the Jurassic through Cenozoic producethe decline in extinction through the Phaner-ozoic recognized by Raup and Sepkoski (1982)and discussed by Gilinsky (1994), that declineis not monotonic. Instead, there are strati-graphically coherent intervals with generallyhigher and lower proportions of extinctionthat are highlighted in Figure 4. Indeed, whenthe Cambrian–Early Ordovician data are re-moved from consideration, it is not obviousthat extinction intensities trend through timefor the remainder of the Phanerozoic; ratherwe see low intensities during the mid Ordo-vician–Early Silurian (except for the late Ash-gillian), Late Carboniferous–mid Permian,and Jurassic to present (except for the end-Cretaceous) separated by discrete intervals ofhigh extinction intensities during the MiddleSilurian–Early Carboniferous and later Perm-ian–Triassic. In total, then, the Phanerozoiccan be divided into six segments based onfluctuation in average extinction proportionper interval (Table 1, Fig. 4). Excluding thepost-Arenig intervals whose extinction ratesfall in a second mode in the lowess/criticalbandwidth analysis described above, the six

532 RICHARD K. BAMBACH ET AL.

segments are apparent by inspection (Fig. 4)and are responsible for the pattern seen in thelowess regression in Figure 5A. Althoughorigination does not have as sharp or distincta set of changes from high to low proportions,the values of proportion of origination gener-ally parallel the pattern for extinction (as willbe seen in Fig. 6).

These six stratigraphically coherent seg-ments provide another ‘‘neighborhood’’ con-text in which variation in origination and ex-tinction may be more meaningfully judged,rather than comparing individual intervalsagainst the full range of values that accumu-lated during the Phanerozoic. Although a lin-ear regression on extinction proportionsthrough the Phanerozoic does have a negativeslope, the decline in extinction rate over timeis caused by three sets of fluctuation betweenconsistently higher and lower rates, with thenegative slope of a linear regression largelydetermined by the anomalously high propor-tions of the Cambrian and Early Ordovicianand the predominantly low rates since the Tri-assic. We make this observation to suggest thatif the pattern of extinction over time is to becharacterized it would be best to model it as athree-phase system, not a monotonic function(nor a two-phase system as suggested by VanValen [1984]).

Using the stratigraphic groupings of inter-vals with similar extinction intensities (Table1, Fig. 4), we see how the outlier intensitiesfrom the critical bandwidth test look withinindividual segments of time. We show this inboth the arithmetic plot (Fig. 4A) and con-verting the proportions to natural logarithms(Fig. 4B). The end-Ordovician (late Ashgillian)and end-Cretaceous (Maastrichtian) pointsare obvious outliers, falling in intervals ofgenerally low proportions of extinction. If thelater Permian points (Guadalupian and Djhul-fian) are regarded as belonging with the restof the Permian (which could be justified be-cause (a) their fauna is taxonomically a contin-uation of the earlier Permian fauna and (b) aswill be noted below, the origination rates forthese two intervals remain in the range of theearlier Permian) they, too, would be extremeoutliers, but if they are grouped with the ele-vated proportions characteristic of the Triassic

they are still clearly higher than any of thoseproportions. However, the late Frasnian (inthe Late Devonian) and the late Norian/Rhaetian (at the end of the Triassic) had ex-tinction proportions that do not appear as out-liers for the larger time segment to which eachbelongs, although the end-Triassic point ishigher than the others in the Triassic.

Three, Not Five, Global Mass Extinctions. Weconclude there is a class of statistically distinctglobal mass extinctions, but that it containsonly three members: the end-Ordovician, end-Permian, and end-Cretaceous. These resultsparallel the conservative bootstrap statisticalanalysis of Hubbard and Gilinsky (1992), whoalso found only these same three unambigu-ous high extinction magnitudes in their anal-ysis.

The other two traditional mass extinctionintervals do not fulfill the criteria for unusualextinction effect. The late Frasnian satisfiesnone of the criteria: it falls within the smoothcontinuum of post-Arenigian extinction mag-nitudes, it is not one of the magnitudes in thesecond mode of post-Arenigian magnitudesusing Silverman’s critical bandwidth test, andit is not an apparent outlier among the mid-Silurian through Early Carboniferous extinc-tion magnitudes. The end-Triassic is some-what more ambiguous because it is one of theintervals in the ‘‘off track’’ high-end tail of therank order distribution of post-Arenigianmagnitudes. However, although it falls at thehigh end of the main mode in Figure 5B, theend-Triassic event does not fall in the separatemode confirmed by the Critical BandwidthTest, nor is it distinctly different from otherTriassic extinction intensities. We do not claimthat the end-Frasnian and end-Triassic wereinnocent of pulsed extinctions, but we do inferthat, by itself, extinction is insufficient to ex-plain the strong diversity depletion at thesetimes.

The Interaction of Originationand Extinction

Paleontologists have focused on mass ex-tinctions as important evolutionary events forthe past quarter of a century or more, but withfew exceptions (e.g., Cutbill and Funnell 1967;Knoll 1989) they have paid less attention to

533ORIGINATION, EXTINCTION, AND DIVERSITY DEPELETIONS

FIGURE 6. Proportions of both genus origination and extinction through the Phanerozoic. Intervals of diversity loss(extinction exceeding origination) are circled. The ‘‘big five’’ mass depletions are numbered.

origination, the other term in the diversityequation. In general, origination has been a fo-cus of interest mostly for the early Paleozoicradiations (Knoll and Carroll 1999; Budd andJensen 2000; Connolly and Miller 2001) andthe recovery intervals following mass extinc-tions (Hart 1996).

The five intervals of diversity loss seen inFigures 1 and 2 have traditionally been re-garded as mass extinctions because extinctionmust have exceeded origination for diversityto drop so markedly during those intervals.However, this focus on extinction ignores twoconditions. One is that there is considerablegenus extinction during any interval of time(over 90% of the intervals of the Phanerozoichave 8% or greater genus extinction), yet whendiversity is maintained because extinction ismatched by origination we are not concernedabout its magnitude. Second is that origina-tion can fluctuate, yet origination is seldomtracked. If origination should fall below typi-cal values while extinction remained at its typ-ical level, diversity would decrease, yet nochange in extinction intensity would have oc-curred. How do the classic ‘‘big five’’ diversitydepletions reflect the interaction betweenorigination and extinction?

Interaction of Origination and Extinction inMajor Diversity Depletions. Figure 6 illus-trates both origination and extinction propor-tions through the Phanerozoic. By definition,

diversity was depleted when extinction wasgreater than origination, and, conversely, di-versity increased when origination exceededextinction. Diversity loss during intervalswhen origination rate was typical for thestratigraphic neighborhood can be ascribed toelevated extinction alone. On the other hand,diversity loss when extinction was not mark-edly higher than the average for its strati-graphic neighborhood must reflect sup-pressed origination—attrition by inadequatereplacement. The intervals during which di-versity decreased are circled on Figure 6 andthe ‘‘big five’’ diversity depletions are num-bered. Using the average proportion of origi-nation and extinction for the larger intervals inwhich each of the ‘‘big five’’ diversity deple-tions falls (Table 1), we can calculate the pro-portional influence (‘‘importance’’) of origi-nation compared with extinction on the loss ofdiversity in each major diversity depletion(Table 2).

During the end-Ordovician, end-Permian,and end-Cretaceous diversity depletions,rates of origination were slightly to markedlygreater than the average for their stratigraphicneighborhoods, whereas extinction rates wereexceptionally high (Fig. 7). By this criterion,then, these three intervals can be regarded astrue global mass extinctions because diversitydepletion in each was driven entirely by ele-vated extinction. Indeed, had rates of origi-

534 RICHARD K. BAMBACH ET AL.

TA

BL

E2.

Act

ual

div

ersi

tych

ang

esd

uri

ng

the

‘‘bi

gfi

ve’’

div

ersi

tyd

eple

tion

san

dth

ech

ang

esth

atw

ould

hav

eoc

curr

edif

orig

inat

ion

orex

tin

ctio

nin

each

inte

rval

had

bee

nth

em

ean

for

the

larg

erse

gm

ent

ofti

me

tow

hic

hea

chis

assi

gn

edin

Tabl

e1.

Cau

seof

div

ersi

tylo

ss

Exc

lusi

vely

elev

ated

exti

nct

ion

Ord

ovic

ian

:la

teA

shg

illi

anP

erm

ian

:D

jhu

lfian

Cre

tace

ous:

Maa

stri

chti

an

Low

orig

inat

ion

plu

sex

tin

ctio

n

Dev

onia

n:

late

Fra

snia

n

Tri

assi

c:la

teN

oria

n/

Rh

aeti

an

Act

ual

dat

aN

o.of

gen

era

ente

rin

gin

terv

alP

rop

orti

onof

gen

us

orig

inat

ion

No.

ofn

ewg

ener

aTo

tal

gen

us

div

ersi

tyin

inte

rval

Pro

por

tion

ofg

enu

sex

tin

ctio

n

1256

0.25

743

516

910.

569

641

0.25

622

086

10.

686

2475

0.20

965

431

290.

474

902

0.17

118

610

880.

347

711

0.21

119

090

10.

468

No.

ofg

ener

ago

ing

exti

nct

No.

ofg

ener

aco

nti

nu

ing

onD

iver

sity

chan

ge

inin

terv

alP

rop

orti

onof

div

ersi

tych

ang

e

962

729

252

72

0.42

0

591

270

371

20.

579

1484

1645

283

02

0.33

5

378

710

219

22

0.21

3

422

479

223

22

0.32

6

Mod

elu

sin

gm

ean

pro

por

tion

ofor

igin

atio

nfo

rap

pro

pri

ate

larg

ein

terv

alin

Tabl

e1

and

actu

alp

rop

orti

onof

exti

nct

ion

Mea

np

rop

orti

onof

orig

inat

ion

sfo

rla

rge

inte

rval

Mod

eln

o.of

new

gen

era

Dif

fere

nce

from

actu

aln

o.of

new

gen

era

Mod

elto

tal

div

ersi

ty

0.24

540

82

2716

64

0.15

711

92

101

760

0.18

556

22

9230

37

0.28

636

117

512

63

0.41

350

031

012

11M

odel

no.

ofex

tin

ctio

ns

(use

real

PE

xt)

Mod

eln

o.of

gen

era

con

tin

uin

gon

Dif

fere

nce

from

actu

alco

nti

nu

ing

(*1)

Mod

eld

iver

sity

chan

ge

inin

terv

alM

odel

pro

por

tion

ald

iver

sity

chan

ge

947

717

212

253

92

0.42

9

521

239

231

240

22

0.62

7

1440

1597

248

287

82

0.35

5

438

825

115

277

20.

085

567

644

165

267

20.

094

Mod

elu

sin

gac

tual

tota

ld

iver

sity

and

mea

np

rop

orti

onof

exti

nct

ion

for

app

rop

riat

ela

rge

inte

rval

inTa

ble

1M

ean

pro

por

tion

ofex

tin

ctio

ns

for

larg

ein

terv

alM

odel

no.

ofex

tin

ctio

ns

Mod

eln

o.of

gen

era

con

tin

uin

gon

Dif

fere

nce

from

actu

alco

nti

nu

ing

(*2)

Mod

eld

iver

sity

chan

ge

inin

terv

alM

odel

pro

por

tion

ald

iver

sity

chan

ge

0.17

429

413

9766

814

10.

112

0.12

811

075

148

111

00.

172

0.12

338

527

4410

9926

90.

109

0.28

731

277

666

212

62

0.14

0

0.34

130

759

411

52

117

20.

165

Ori

gin

atio

ns:

rati

ore

al/

mea

nE

xtin

ctio

ns:

rati

ore

al/

mea

n1.

066

3.27

21.

849

5.37

31.

164

3.85

50.

515

1.21

20.

380

1.37

5

Pro

por

tion

alin

flu

ence

ofor

igin

atio

n(*

1/*1

1*2

)P

rop

orti

onal

infl

uen

ceof

exti

nct

ion

(*2/

*11

*2)

20.

018

1.01

82

0.26

61.

266

20.

046

1.04

60.

635

0.36

50.

589

0.41

1

535ORIGINATION, EXTINCTION, AND DIVERSITY DEPELETIONS

FIGURE 7. Relationship between origination and extinction for the three mass diversity depletions that are truemass extinctions. Arrow marked O indicates the difference between the mean origination for the larger interval towhich each is assigned (Table 1), shown by a dashed horizontal line, and the actual origination in the mass depletioninterval. Arrow marked E indicates the difference between the mean extinction for the larger interval to which eachis assigned (Table 1), shown by a horizontal solid line, and the actual extinction magnitude in the mass depletioninterval. Diversity loss indicated by the vertical thin box between the origination and extinction values. A, LateAshgillian, at the end of the Ordovician. B, Djhulfian, at the end of the Permian (with marked diversity loss in thepreceding Guadalupian stage, as well). C, Maastrichtian, at the end of the Cretaceous.

nation during these intervals been no morethan average, diversity losses would actuallyhave been somewhat greater than they were(Table 2). The noticeably higher-than-averageproportion of origination in the Djhulfian(10% higher than the mean for the Late Car-boniferous and Permian) may reflect recoveryfrom the diversity loss in the Guadalupian,but the other two intervals have proportionsof origination only one to two-and-a-half per-cent higher than the average for their strati-graphic neighborhoods.

In contrast, the two remaining ‘‘big five’’mass depletions—the late Frasnian and theend-Triassic—reflect a more complicated im-balance between origination and extinction(Fig. 8). During both intervals, origination wasmarkedly lower than average for their strati-graphic neighborhoods. Although each inter-val registered somewhat elevated extinction,about two-thirds of the diversity loss in thelate Frasnian and almost 60% in the end-Tri-assic can be ascribed to origination failure, notelevated extinction (Table 2). McGhee (1988)originally emphasized that low originationwas a major factor in the Frasnian diversitydepletion, but it is only with this paper that

the role of origination failure is brought to thefore for both the Frasnian and the end-Trias-sic. Neither was an extinction event of themagnitude of the end-Ordovician, end-Perm-ian, and end–Cretaceous events, yet each qual-ifies as one of the ‘‘big five’’ post-Cambrian di-versity depletions because of moderately ele-vated extinction in concert with marked orig-ination failure.

Other intervals that have relatively highmagnitude of extinction are occasionally la-beled ‘‘mass extinctions,’’ but they do notshow dramatic losses of diversity. For exam-ple, the Kacak/otomari event near the end ofthe Eifelian in the Middle Devonian has beenassociated with impact ejecta in Morocco (Ell-wood et al. 2003), but the cited extinction of‘‘as many as 40% of all marine animal genera’’is actually 37% extinction for the entire 11-mil-lion-year-long Eifelian stage, not just the Ka-cak/otomari event (Sepkoski’s data, cited byEllwood et al. is only resolved to the full stagelevel for the Eifelian), and is nearly balancedby 33% origination, resulting in a net Eifeliandiversity loss of only 4%. Extinction was equalto the Eifelian level in the preceding Emsianand was actually greater in the Ludlovian of

536 RICHARD K. BAMBACH ET AL.

FIGURE 8. Relationship between origination and extinction for the two mass diversity depletions that are not ex-clusively extinction driven. Arrows as in Figure 7. Diversity loss indicated by the vertical thin box between theorigination and extinction values. A, Late Frasnian, in the Late Devonian. Note that several preceding intervals alsohave lower origination than extinction, indicating continuous diversity loss, although a drop in origination plus asmall peak of extinction concentrates more diversity loss in the late Frasnian. B, Late Norian/Rhaetian, at the endof the Triassic. In this case, proportion of origination decreased through most of the Triassic as diversity reboundedfrom the end-Permian mass depletion, but exceeded extinction except at the end of the Triassic, when originationwas at its low point and there was a small peak of extinction.

the Silurian, but in both instances originationalso nearly balanced extinction. Events thatrecord moderate extinction over short time in-tervals in some regions, like the Kacak/oto-mari episode, deserve close attention, but theyare simply not comparable to the end-Permianor end-Cretaceous devastations, nor were theyassociated with diversity depletions on a glob-al scale similar to those of the late Frasnian orend-Triassic.

Mass Depletions during the Cambrian. TheCambrian and Early Ordovician are unusualin having very low diversity and very highrates of faunal turnover. Figures 3A, 4, and 9Billustrate the preponderance of high propor-tional origination and extinction in this inter-val: 87% of the intervals in the Phanerozoicwith genus origination greater than 50% and68% of all intervals with genus extinctiongreater than 40% occur during the Cambrianand Early Ordovician. As noted above, it isstatistically highly unlikely that the prepon-derance of high extinction magnitude wasproduced by chance alone. What makes thiseven more unusual is that the durations ofmany subdivisions during this interval aremuch shorter than the average through the

rest of the Phanerozoic, yet taxonomic turn-over is greater (as can be seen by the relativesize of the turnover peaks compared with theboundary-crossing [minimum] diversity plotin Figs. 1 and 9A). Sepkoski’s (1996) strati-graphic subdivisions of Cambrian time maybe unrealistically small (e.g., Landing 1994),raising questions about the accuracy of his de-tailed correlations and calling his diversitynumbers for these fine-scale intervals intoquestion; however, combining these intervalsinto larger time divisions more comparable induration to those characteristic of younger pe-riods would result in even more pronouncedturnover rates because of the short ranges ofmany Cambrian taxa. As an example, propor-tion of origination in the Middle Cambrian asa whole was 0.691 and proportion of extinc-tion was 0.676, whereas origination averaged0.461 and extinction averaged 0.467 for each ofthe four intervals of the Middle Cambrian tab-ulated here. Thus, regardless of any over-op-timism in the designation of stratigraphic binsused by Sepkoski, the Cambrian and Early Or-dovician was an interval of unusual turnover.

Important extinctions have, of course, beenrecognized in the Cambrian. Signor (1992)

537ORIGINATION, EXTINCTION, AND DIVERSITY DEPELETIONS

FIGURE 9. A, Genus diversity and diversity turnover through the Cambrian and Early Ordovician. Graph plottedas in Figure 1. Notice the high degree of taxonomic turnover. Arrows mark the two times of marked diversity de-crease. B, Proportions of origination and extinction through the Cambrian and Early Ordovician. Both originationand extinction were generally high during the Cambrian and Early Ordovician. The two intervals of marked di-versity depletion are distinguished by depression of origination, not peaks of extinction.

first pointed to the Botomian as a time of massextinction (see also Zhuravlev and Wood1996). Archaeocyathan diversity collapsed,and small shelly fossil diversity declined aswell (although, as Porter [2003] has shown,loss of preservational opportunity clouds thepicture for the latter group). Nonetheless, Sep-koski’s data suggest that Botomian extinctionproportions for the total fauna were actuallylower than those of several other Cambrian in-tervals, and the marked late Botomian changefor the fauna as a whole reflects a drop in orig-ination, not an increase in extinction (Fig. 9B).The loss of diversity in the late Botomian wasnot because extinction proportion for arcaheo-

cyathids increased, but rather because propor-tions of origination for archaeocyathids, prob-lematica, hyolithids, and inarticulate brachio-pods all fell to one-fifth to one-third the valuesthey had maintained in the Atdabanian andearly Botomian. Among diverse groups in theEarly Cambrian, only the trilobites main-tained origination rates high enough to coun-ter attrition from the ‘‘normal’’ extinction pro-portions in the late Botomian.

The same is true for the early part of the lateMiddle Cambrian. Although Palmer’s (1984,1998) documentation of extinction at biomereboundaries may contribute to the high extinc-tion magnitudes in the Middle and Late Cam-

538 RICHARD K. BAMBACH ET AL.

brian, the diversity loss of the early part of thelate Middle Cambrian interval stands outmostly for its low level of genus origination(Fig. 9B). Thus, the two Cambrian intervals ofdiversity depletion resemble the Late Devo-nian and end-Triassic more closely than theydo the three unambiguous global mass extinc-tions.

In one way, this finding is more or less in-evitable. When mean rates of extinction standat 46% per interval, driving diversity deple-tion by increasing extinction rate is a tall order.However, the more specific question of whyarchaeocyathans, trilobites, and other Cam-brian dominants should have experienced de-clining rates of origination during these inter-vals has no clear answer at this time (as is alsothe case for the Frasnian and end-Triassic not-ed above). Perhaps changing Botomian oceanchemistry selected against the massive calci-fication that is the hallmark of archaeocy-athans (e.g., Rowland and Shapiro 2002). Andperhaps ocean circulation changed during thelate Middle Cambrian, influencing the shelfenvironments that supported marine diversi-ty. Compelling answers may come from cre-ative interactions between paleontology andEarth system geochemistry.

What about Continuity of Effectand Cause?

Although the ‘‘big five’’ global mass deple-tions of diversity do not share continuity ofmagnitude it might be that they share com-mon properties in terms of the biological ef-fects of diversity loss. This, however, does notappear to be the case.

Knoll et al. (1996) identified a distinct selec-tivity in end-Permian extinction in whichheavily calcified animals with low metabolicrates were devastated (88% genus extinction),whereas metazoans with high metabolic ratesand calcareous skeletons precipitated fromphysiologically buffered solutions (or withskeletons made from materials other thanCaCO3) were far less affected (49% genus ex-tinction), and more recent work is refining thepattern (Bambach and Knoll unpublisheddata). During the end-Ordovician and end-Cretaceous events, no selectivity of this sortoccurred. Sheehan et al. (1996) documented

selectivity differences in the patterns of ex-tinction for these two events that clearly dif-ferentiate between them, and neither resem-bles the Permian pattern. Thus, although theend-Permian and end-Cretaceous global massextinctions share one result—both devastatedantecedent ecosystems—the patterns of thetwo extinctions were different and their post-extinction recoveries distinct (Bambach et al.2002). As pointed out by Droser et al. (2000),Brenchley et al. (2001), and Sheehan (2001),and confirmed quantitatively by Bambach etal. (2002), the Ordovician global mass extinc-tion did not have a comparable effect on eco-system structure.

As an illustration of the difference in effectof the Ordovician and Permian extinctions, arecent study of brachiopods in South China(Rong and Shen 2002) noted that there wereno ordinal extinctions in the Ordovician eventand that there is a ‘‘close taxonomic and eco-logical relationship of pre- and post-extinctionbrachiopod associations’’ (p. 25). In contrast,four orders and 20 superfamilies disappearedin the end-Permian event, and very few Tri-assic brachiopods link back to the Permian atthe genus level. For the three unambiguousglobal mass extinctions, then, there are differ-ences in the physiological selectivity of theevents, the ecological impacts of the events,and the style of effect on particular taxa. Con-tinuity of effect is simply absent.

The two other global mass depletions differfrom the three global mass extinctions in nu-merous ways, including effect when the bal-ance between origination and extinction isconsidered, but they are still not well under-stood. The Frasnian event is complicated.Most workers accept that at least two pulses ofextinction occurred, in association with twoKellwasser black-shale horizons (see, for ex-ample, Chen and Tucker 2003). Extinctionpulses are also recorded in the Middle andLate Devonian (House 1985; McGhee 1996),especially the so-called Hangenberg event atthe end of the Famennian (House 1985; Si-makov 1993; Wang et al. 1993). Indeed, at thestage and substage levels, marine diversitydecreased almost continuously through Mid-dle and Late Devonian time (Fig. 1). Moreover,the marine fauna underwent dramatic turn-

539ORIGINATION, EXTINCTION, AND DIVERSITY DEPELETIONS

over in dominant organisms (Bambach 1999).Copper (2002: p. 49) observed that reef distri-bution declined progressively from the Give-tian through the Frasnian, and stated thatthere is ‘‘no catastrophic reef ‘kill horizon’ atthe F/F boundary known anywhere in theworld.’’ Copper (2002) associated the ob-served decline with climatic and tectonicchange. Although the causes of reduced Fras-nian origination are not yet clearly known,loss of complex reef habitats could have facil-itated reduced origination by eliminating anormally diverse ecosystem.

The end-Triassic event is the least well un-derstood of the major diversity depletions. Asnoted above, this event was driven as much byorigination failure as by extinction, in contrastto the end-Ordovician, end-Permian, and end-Cretaceous events. Moreover, neither glacia-tion nor major biotic reorganization, charac-teristics associated with the end-Ordovicianand Late Devonian diversity depletions, ismanifest in the Late Triassic (although bothlarge plateau basalt eruptions and bolide im-pact have been implicated [Bice et al. 1992;Courtillot et al. 1996; Wignall 2001; Olsen etal. 2002]). Although complete and informativemarine sections are still being sought (Hallamet al. 2000), there are now proposals that ob-served terrestrial and marine extinctions maynot be synchronous (Palfy et al. 2000) and geo-chemical analyses that also suggest a double-pulsed event of some sort (Hesselbo et al.2002). In concluding a recent survey of end-Triassic data, Hallam (2002: p. 133) stated,‘‘Though the current evidence cannot deci-sively exclude it, the foregoing review doesnot favour a major T–J boundary catastrophe.’’

The five major mass depletions of diversityare so different in major features that it is clearthey cannot share common cause. Nor didthey have common effect, beyond major lossof diversity.

Could It All Be Artifactual?

Variability in the quality of the geologic re-cord has the potential to affect observed di-versity and it undoubtedly has affected someaspects of detailed interval-to-interval chang-es in apparent diversity (Peters and Foote2001, 2002; Smith et al. 2001; Kier 1974 1977).

One might, therefore, question whether ourconclusions reflect artifactual, rather than bi-ological, patterns in the fossil record. Couldthere be defects or biases in the record thatgive the appearance of major diversity deple-tions when none actually exist?

Although we do not deny that some of theapparent diversity change in each of the ‘‘bigfive’’ diversity depletions could be accentuat-ed by record failure, in none of these instancesis it reasonable to expect that tabulated diver-sity change was produced entirely by failure ofthe rock record. For instance, Raup (1978) onceestimated the amount of extinction at the endof the Permian was equivalent to an 85-mil-lion-year gap in the geologic record, some-thing obviously precluded by radiometricages of Permian and Triassic rocks near thePermo-Triassic boundary (Bowring et al.1998). Also, Raup’s (1982) efforts to model ex-tinction by elimination of large areas of theEarth—originally conceived as a test of im-pact scenarios for extinction, but equally ap-plicable to questions of record failure—pro-duced surprisingly small extinctions of generaand families (mean extinction of only 13.8% ofmarine genera and 12% of terrestrial families)when half our planet’s surface was eradicated.Record failure is simply not capable of pro-ducing apparent diversity fluctuations on thescale observed in the large global diversity de-pletions.

In summary, we believe that the major con-clusions articulated in this paper describe realchanges in the diversity of marine animalsthough Phanerozoic time.

Conclusions

Of 88 stage or substage intervals during the470 million years since the Early OrdovicianArenig, only five suffered a loss of diversitygreater than 13.5% of the number of marinegenera that entered the interval. These five in-tervals—the late Ashgillian at the end of theOrdovician, the late Frasnian in the Late De-vonian, the Djhulfian at the end of the Perm-ian, the late Norian/Rhaetian at the end of theTriassic, and the Maastrichtian at the end ofthe Cretaceous—have been grouped as the‘‘big five’’ mass extinctions, but they are betterregarded as mass depletions of diversity.

540 RICHARD K. BAMBACH ET AL.

Two of the five mass depletions, the lateFrasnian and the end-Triassic, resulted pri-marily from attrition related to reduced orig-ination, rather than devastation from unusu-ally elevated extinction. Both intervals oc-curred during times of generally high extinc-tion, so lowered origination had severeconsequences for diversity. Extinction in thesetwo intervals was elevated above average fortheir stratigraphic neighborhoods, but in eachcase elevated extinction was responsible foronly a little more than one-third of the diver-sity loss, leaving the remainder to be ex-plained by reduced origination. A similar re-lationship between reduced origination anddiversity loss is seen in two Cambrian inter-vals with marked loss of diversity (the late Bo-tomian and the early part of the late MiddleCambrian).

On the other hand, diversity loss in the end-Ordovician, end-Permian, and end-Creta-ceous intervals resulted exclusively from ele-vated extinction. The extinction magnitudes ofthese three intervals are statistically differentfrom the extinction magnitudes in their strati-graphic neighborhoods; they can be regardedas ‘‘true’’ global mass extinctions.

Acknowledgments

We thank A. I. Miller and an anonymous re-viewer for many suggestions that helped toimprove the manuscript. A.H.K.’s researchwas supported, in part, by the NASA Astro-biology Institute. S.C.W. is grateful for fund-ing from the Swarthmore College ResearchFund and the Christian R. and Mary F. Lind-back Foundation.

Literature CitedBambach, R. K. 1983. Ecospace utilization and guilds in marine

communities through the Phanerozoic. Pp. 719–746 (Chapter15) in M. Tevesz and P. McCall, eds. Biotic interactions in re-cent and fossil benthic communities. Plenum, New York.

Bambach, R. K. 1985. Classes and adaptive variety: the ecologyof diversification in marine faunas through the Phanerozoic.Pp. 191–253 in J. W. Valentine, ed. Phanerozoic diversity pat-terns: profiles in macroevolution. Princeton University Press,Princeton, N.J.

Bambach, R. K. 1999. Energetics in the global marine fauna: aconnection between terrestrial diversification and change inthe marine biosphere. Geobios 32:131–144.

Bambach, R. K., and N. L. Gilinsky. 1986. Perspectives on thedistribution of origination and extinction during the Phan-erozoic. Geological Society of America Abstracts with Pro-grams 18:534.

Bambach, R. K., A. H. Knoll, and J. J. Sepkoski Jr. 2002. Anatom-ical and ecological constraints on Phanerozoic animal diver-sity in the marine realm. Proceedings of the National Acad-emy of Sciences USA 99:6854–6859.

Bice, D. M., C. R. Newton, S. McCauley, P. W. Reiners, and C. A.McRoberts. 1992. Shocked quartz at the Triassic-Jurassicboundary in Italy. Science 255:443–446.

Bowring, S. A., D. H. Erwin, Y. G. Jin, M. W. Martin, K. Davidek,and W. Wang. 1998. U/Pb zircon geochronology and tempoof the end-Permian mass extinction. Science 280:1039–1045.

Brasier, M. D., and S. S. Sukhov. 1998. The falling amplitude ofcarbon isotopic oscillations through the lower to middle Cam-brian: northern Siberia data. Canadian Journal of Earth Sci-ences 35:353–373.

Brenchley, P. J., J. D. Marshall, and C. J. Underwood. 2001. Doall mass extinctions represent an ecological crisis?’’ Evidencefrom the Late Ordovician. Geological Journal 36:329–340.

Budd, G. E., and S. Jensen. 2000. A critical reappraisal of thefossil record of the bilaterian phyla. Biological Reviews 75:253–295.

Bush, A. M., and R. K. Bambach. In press. Did alpha diversityincrease during the Phanerozoic? Lifting the veil of tapho-nomic, latitudinal, and environmental biases in the study ofpaleocommunities. Journal of Geology.

Bush, A. M., M. J. Markey, and C. R. Marshall. 2001. Estimatingbiodiversity in light of heterogeneous sampling intensities:the relationship between taxon-occurrence subsamplingmethods and alpha and beta diversity. PaleoBios 21(Suppl. toNo. 2):37.

Bush, A. M., M. J. Markey, and C. R. Marshall. In press. Alpha,beta, gamma: the effects of spatially organized biodiversityon sampling-standardization. Paleobiology.

Chen, D., and M. E. Tucker. 2003. The Frasnian-Famennian massextinction: insights from high-resolution sequence stratigra-phy and cyclostratigraphy in South China. Palaeogeography,Palaeoclimatology, Palaeoecology 193:87–111.

Cleveland, W. S. 1979. Robust locally weighted regression andsmoothing scatterplots. Journal of the American StatisticalAssociation 74:829–836.

Connolly, S. R., and A. I. Miller. 2001. Global Ordovician faunaltransitions in the marine benthos: proximate causes. Paleo-biology 27:779–795.

Copper, P. 2002. Reef development at the Frasnian/Famennianmass extinction boundary. Palaeogeography, Palaeoclimatol-ogy, Palaeoecology 181:27–65.

Courtillot, V. J.-J. Jaeger, Z. Yang, G. Feraud, and C. Hofmann.1996. The influence of continental flood basalts on mass ex-tinctions: where do we stand? In G. Ryder, D. Fastovsky, andS. Gartner, eds. The Cretaceous-Tertiary event and other ca-tastrophes in earth history. Geological Society of AmericaSpecial Paper 307:513–515.

Cutbill, J. L., and B. M. Funnell, 1967. Numerical analysis of thefossil record. Pp. 791–820 in W. B. Harland, W. B., C. H. Hol-land, M. R. House, N. F. Hughes, A. B. Reynolds, M. J. S. Rud-wick, G. E. Satterthwaite, L. B. Tarlow, and E. C. Willey, eds.The fossil record. Geological Society, London.

Droser, M. L., D. J. Bottjer, P. M. Sheehan, and G. R. McGhee.2000. Decoupling of taxonomic and ecologic severity of Phan-erozoic marine mass extinctions. Geology 28:675–678.

Ellwood, B. B., S. L. Benoist, A. El Hassani, C. Wheeler, and R.E. Crick. 2003. Impact ejecta layer from the Mid-Devonian:possible connection to global mass extinctions. Science 300:1734–1737.

Finney, S. C., W. B. N. Berry, J. D. Cooper, R. L. Ripperdan, W.C. Sweet, S. R. Jacobson, A. Soufiane, A. Achab, and P. J. Noble.1999. Late Ordovician mass extinction: a new perspectivefrom stratigraphic sections in central Nevada. Geology 27:215–218.

541ORIGINATION, EXTINCTION, AND DIVERSITY DEPELETIONS

Foote, M. 1988. Survivorship analysis of Cambrian and Ordo-vician trilobites. Paleobiology 14:258–271.

Gilinsky, N. L. 1994. Volatility and the Phanerozoic decline ofbackground extinction intensity. Paleobiology 20:445–458.

Hallam, A. 2002. How catastrophic was the end-Triassic massextinction? Lethaia 35:147–157.

Hallam, A., and P. B. Wignall. 1999. Mass extinctions and sea-level changes. Earth-Science Reviews 48:217–250.

Hart, M. B., ed. 1996. Biotic recovery from mass extinctionevents. Geological Society of London Special Publication 102.

Hesselbo, S. P., S. A. Robinson, F. Surlyk, and S. Piasecki. 2002.Terrestrial and marine extinction at the Triassic-Jurassicboundary synchronized with major carbon-cycle perturba-tion: a link to initiation of massive volcanism? Geology 30:251–254.

House, M. R. 1985. Correlation of mid Paleozoic ammonoid evo-lutionary events with global sedimentary perturbations. Na-ture 313:17–22.

Hubbard, A. E., and N. L. Gilinsky. 1992. Mass extinctions asstatistical phenomena: an examination of the evidence usingx2 tests and bootstrapping. Paleobiology 18:148–160.

Jablonski, D. J. 1986. Causes and consequences of mass extinc-tions: a comparative approach. Pp. 183–229 in D. K. Elliott, ed.Dynamics of extinction. Wiley, New York.

Jablonski, D. J., K. Roy, J. W. Valentine, R. M. Price, and P. S. An-derson. 2003. The impact of the Pull of the Recent on the his-tory of marine diversity. Science 300:1133–1135.

Jackson, J. B. C., and K. G. Johnson. 2001. Measuring past bio-diversity. Science 293:2401–2404.

Jin, Y.-g., Y. Wang, W. Wang, Q. H. Shang, C. Q. Cao, and D. H.Erwin. 2000. Pattern of marine mass extinction near the Perm-ian/Triassic boundary in South China. Science 289:432–436.

Kier, P. M. 1974. Evolutionary trends and their functional sig-nificance in post-Paleozoic echinoids. Paleontological SocietyMemoir 5. Journal of Paleontology 48(Suppl. to No. 3).

———. 1977. The poor fossil record of the regular echinoid. Pa-leobiology 3:168–174.

Knoll, A. H. 1989. Evolution and extinction in the marine realm:some constraints imposed by phytoplankton. PhilosophicalTransactions of the Royal Society of London B 325:279–290.

Knoll, A. H., R. K. Bambach, D. E. Canfield, and J. P. Grotzinger.1996. Comparative earth history and Late Permian mass ex-tinction. Science 273:452–457.

Knoll, A. H., and S. B. Carroll. 1999. Early animal evolution:emerging views from comparative biology and geology. Sci-ence 284:2129–2137.

Landing, E. 1994. Lower Cambrian of southeastern Newfound-land: epeirogeny and Lazarus faunas, lithofacies-biofacieslinks, and the myth of a global chronostratigraphy. Pp. 283–309 in J. H. Lipps and P. W. Signor, eds. Origin and early evo-lution of the Metazoa. Plenum, New York.

Leakey, R. E., and R. Lewin. 1995. The sixth extinction: patternsof life and the future of humankind. Doubleday, New York,271 p.

MacLeod, N., and G. Keller, eds. 1996. Cretaceous-Tertiary massextinctions: biotic and environmental changes. Norton, NewYork.

McGhee, George R., Jr. 1988. The Late Devonian extinctionevent: evidence for abrupt ecosystem collapse. Paleobiology14:250–257.

———. 1996. The Late Devonian mass extinction. Columbia Uni-versity Press, New York.

———. 2001. The ‘multiple impacts hypothesis’ for mass extinc-tion: a comparison of the Late Devonian and the late Eocene.Palaeogeography, Palaeoclimatology, Palaeoecology 176:47–58.

Newell, N. D. 1962. Paleontological gaps and geochronology.Journal of Paleontology 36:592–610.

———. 1963. Crises in the history of life. Scientific American208(2):77–92.

———. 1967. Revolutions in the history of life. In C. C. AlbrittonJr., ed. Uniformity and simplicity. Geological Society of Amer-ica Special Paper 89:63–91.

Olsen, P. E., D. V. Kent, H. D. Sues, C. Koeberl, H. Huber, A.Montanari, E. C. Rainforth, S. J. Fowell, M. J. Szajna, and B.W. Hartline. 2002. Ascent of dinosaurs linked to an iridiumanomaly at the Triassic-Jurassic boundary. Science 296:1305–1307.

Palfy, J., J. K. Mortensen, E. S. Carter, P. L. Smith, R. M. Fried-man, and H. W. Tipper. 2000. Timing the end-Triassic massextinction: first on land, then in the sea? Geology 28:39–42.

Palmer, A. R. 1984. The biomere problem: evolution of an idea.Journal of Paleontology 58:599–611.

———. 1998. A proposed nomenclature for stages and series forthe Cambrian of Laurentia. Canadian Journal of Earth Scienc-es 35:323–328.

Peters, S. E., and M. Foote. 2001. Biodiversity in the Phanero-zoic: a reinterpretation. Paleobiology 27:583–601.

———. 2002. Determinants of extinction in the fossil record. Na-ture 416:420–424.

Porter, S. M. 2004. Closing the ‘‘phosphatization’’ window: test-ing for the influence of taphonomic megabias on the patternof small shelly fossil decline. Palaios 19:178–183.

Quinn, J. F. 1983. Comment on: mass extinctions in the fossil re-cord. Science 219:1239–1240.

Raup, D. M. 1978. Cohort analysis of generic survivorship. Pa-leobiology 4:1–15.

———. 1982. Biogeographic extinction: a feasibility test. In L. E.Silver and P. H. Schultz, eds. Geological implications of im-pacts of large asteroids and comets on the Earth. GeologicalSociety of America Special Paper 190:277–281.

———. 1987. Mass extinction: a commentary. Palaeontology 30:1–13.

———. 1991. A kill curve for Phanerozoic marine species. Pa-leobiology 17:37–48.

Raup, D. M., and J. J. Sepkoski Jr. 1982. Mass extinctions in themarine fossil record. Science 215:1501–1503.

Rong, J.-y., and S.-z. Shen. 2002. Comparative analysis of theend-Permian and end-Ordovician brachiopod mass extinc-tions and survivals in South China. Palaeogeography, Palaeo-climatology, Palaeoecology 188:25–38.

Rowland, S. M., and R. S. Shapiro. 2002. Reef patterns and en-vironmental influences in the Cambrian and earliest Ordo-vician. In W. Kiessling, E. Flugel, and J. Golonka, eds. Phan-erozoic reef patterns. SEPM Special Publication. 72:95–128.

Saltzman, M. R., R. L. Ripperdan, M. D. Brasier, K. C. Lohmann,R. A. Robison, W. T. Chang, S. C. Peng, E. K. Ergaliev, and B.Runnegar. 2000. A global carbon isotope excursion (SPICE)during the Late Cambrian: relation to trilobite extinctions, or-ganic-matter burial and sea level. Palaeogeography, Palaeo-climatology, Palaeoecology 162:211–223.

Sepkoski, J. J., Jr. 1981. A factor analytic description of the Phan-erozoic marine fossil record. Paleobiology 7:36–53.

———. 1982. A compendium of fossil marine families. Milwau-kee Public Museum Contributions to Biology and GeologyNo. 51.

———. 2002. A compendium of fossil marine animal genera.Bulletins of American Paleontology No. 363.

Sepkoski, J. J., Jr., R. K. Bambach, D. M. Raup, and J. W. Valentine.1981. Phanerozoic marine diversity: a strong signal from thefossil record. Nature 293:435–437.

Sheehan, P. M. 2001. The Late Ordovician mass extinction. An-nual Reviews of Earth and Planetary Science 29:331–364.

Sheehan, P. M., P. J. Coorough, and D. E. Fastovsky. 1996. Bioticselectivity during the K/T and Late Ordovician extinctionevents. In G. Ryder, D. Fastovsky, and S. Gartner, eds. The

542 RICHARD K. BAMBACH ET AL.

Cretaceous-Tertiary event and other catastrophes in earth his-tory. Geological Society of America Special Paper 307:477–489.

Signor, P. W. 1992. Taxonomic diversity and faunal turnover inthe Early Cambrian: did the most severe mass extinction ofthe Phanerozoic occur in the Botomian stage? In S. Lidgardand P. R. Crane, eds. Fifth North American PaleontologicalConvention, Abstracts and program. Paleontological SocietySpecial Publication 6:272.

Signor, P. W., and J. H. Lipps. 1982. Sampling bias, gradual ex-tinction patterns, and catastrophes in the fossil record. In L.E. Silver and P. H. Schultz, eds. Geological implications of im-pacts of large asteroids and comets on the Earth. GeologicalSociety of America Special Paper 190:291–296.

Silverman, B. W. 1981. Using kernel density estimates to inves-tigate multimodality. Journal of the Royal Statistical Society B43:97–99.

Simakov, K. V. 1993. Biochronological aspects of the Devonian-Carboniferous crisis in the regions of the former U.S.S.R. Pa-

laeogeography, Palaeoclimatology, Palaeoecology 104:127–137.

Smith, A. B., A. S. Gale, and N. E. A. Monks. 2001. Sea-levelchange and rock-record bias in the Cretaceous: a problem forextinction and biodiversity studies. Paleobiology 27:241–253.

Stanley, S. M., and X. Yang. 1994. A double mass extinction atthe end of the Paleozoic era. Science 266:1340–1344.

Van Valen, L. M. 1984. A resetting of Phanerozoic communityevolution. Nature 307:50–52.

Wang, K., M. Attrep, and C. J. Orth. 1993. Global iridium anom-aly, mass extinction, and redox change at the Devonian-Car-boniferous boundary. Geology 21:1071–1074.

Wang, S. C. 2003. On the continuity of background and massextinction. Paleobiology 29:455–467.

Wignall, P. B. 2001. Large igneous provinces and mass extinc-tions. Earth-Science Reviews 53:1–33.

Wignall, P. B., and R. J. Twitchett. 1996. Oceanic anoxia and theend Permian mass extinction. Science 272:1155–1158.

Zhuravlev, A. Yu., and R. A. Wood. 1996. Anoxia as the causeof the mid-Early Cambrian (Botomian) extinction event. Ge-ology 24:311–314.

4 MARCH 2005, GSA TODAY

ABSTRACTRightly or wrongly, dinosaurs are

poster children for the Cretaceous-Tertiary (K-T) extinction. The rate and cause of their extinction, however, has been contentious, at least in part because of their rarity. Nonetheless, significant data have accumulated to indicate that the dinosaur extinction, in North America at least, was geo-logically instantaneous. The evidence comes from field studies in geologi-cally disparate settings involving the reconstruction of dinosaur stratigraphic ranges as well as community structure in the Late Cretaceous, and from quan-titative studies of the post-Cretaceous evolution of mammals.

The hypothesis of extinction by aster-oid impact is concordant with what is known of the rate of the dinosaur extinction, as well as the patterns of selective vertebrate survivorship across the K-T boundary. The precise nature of the kill mechanism(s), however, remains under discussion.

INTRODUCTIONThe question of what happened to

the dinosaurs at the Cretaceous-Tertiary (K-T) boundary has come to exem-plify the K-T extinction. Did they die out instantly, or were they gradually going extinct over millions of years? As has been noted by Clemens et al. (1981), identifying the patterns of the dinosaur extinction is a question quite separate from, but a prerequisite to, identifying the cause(s) of the extinc-tion itself. In the last 25 years, much has been learned about the patterns of the dinosaur extinction. Ultimately, what we know about extinction pat-

terns constrains causal mechanisms, a point forcibly made by Bakker (1986). Here, we review a variety of different studies, all of which ultimately converge on the conclusion that the extinction of the dinosaurs in North America was geologically instantaneous. From this conclusion and data pertaining to the post-Cretaceous recovery, we consider potential causes of the extinction.

For many years it has been said that dinosaurs were waning in num-ber and diversity over the last 10 m.y. of the Cretaceous (the Campanian-Maastrichtian interval). A typical state-ment of this viewpoint can be found in Dodson (1996) who notes, “[Among dinosaurs] I see a pattern of dwindling. Ten million years before the end [the K-T boundary] there were two subfamilies of ceratopsids. …At the end, only the chasmosaurines were left. Ten million years before the end, there were two families of hadrosaurs. …At the end, only the hadrosaurines were left. Ten million years before the end, there were two families of armoured dinosaurs. …At the end, only the ankylosaurids were left” (p. 280).

This apparent drop in diversity looks to us to be comparable to other Late Cretaceous fluctuations in the imperfect dinosaur record. Considered in the con-text of all dinosaur diversity fluctuations throughout the Late Cretaceous, this drop in diversity over the last 10 m.y. does not appear remarkable, either for North America or globally (Fastovsky et al., 2004). Consequently, we focus here on the final two m.y. of the dinosaur record as key to the rate and mecha-nism of their extinction.

It would be ideal to be able to resolve the precise duration of the North American dinosaur extinction, no matter what its length. At a tem-poral distance of 65 m.y. and beset by a fragmentary terrestrial record, however, we can only characterize events as geologically instantaneous, by which we mean encompassing time-scales of tens of thousands of years (or less). Nonetheless, this allows us to distinguish between processes and events that occurred on such time-scales (or less) and those that occurred on longer ones.

Global databases for dinosaurs exist (e.g., Weishampel et al., 2004), and fluxes in dinosaur diversity have been reconstructed from them (e.g., Dodson, 1990; Fastovsky et al., 2004); yet, the North American record remains uniquely suited to understanding the rate of the dinosaur extinction. This is because only in North America are there dinosaur-bearing exposures with a high level of stratigraphic resolution that preserve a terrestrial K-T boundary and that have been studied quantitatively.

SEDIMENTARY ENVIRONMENTS THAT PRESERVE THE LATE HISTORY OF THE DINOSAURS

In the latest Cretaceous of the North American Western Interior, dinosaurs such as Triceratops, Tyrannosaurus, and Edmontosaurus (and a host of lesser luminaries) roamed upland and coastal plain settings (Lehman, 1987) that formed during the Laramide phase of the Rocky Mountain uplift (Peterson, 1986). Dinosaur-bearing units that have been the subjects of studies suf-ficiently detailed to resolve the nature of the extinction are preserved in the structurally complicated Hanna Basin, an intermontane basin in southern Wyoming (Eberle and Lillegraven, 1998; Lillegraven et al., 2004), and unde-formed sediments of the Williston Basin, an intracratonic basin extending through eastern Montana and western North and South Dakota (Peterson, 1986) (Fig. 1).

In the Hanna Basin, the K-T bound-ary is found within the Ferris Formation,

GSA Today: v. 15, no. 3, doi: 10:1130/1052-5173(2005)015<4:TEOTDI>2.0.CO;2

The Extinction of the Dinosaurs in North AmericaDavid E. Fastovsky*, Department of Geosciences, University of Rhode Island, 9 East

Alumni Ave., Kingston, Rhode Island 02881, USA, defastov@uri.edu, and Peter M.

Sheehan, Department of Geology, Milwaukee Public Museum, 800 West Wells

Street, Milwaukee, Wisconsin 53233, USA, sheehan@mpm.edu

*Present address (2004–2005): Instituto de Geología, Universidad Nacional Autónoma de México, Ciudad Universitaria, Coyaocan 04510, México D.F., México.

GSA TODAY, MARCH 2005 5

a 1.2-km-thick sequence of sandstones and mudstones inter-preted to represent fluvial deposition. A well-developed braided river system bisects the Cretaceous part of the for-mation, giving way (~300 m below the K-T boundary) to a meandering fluvial system with lacustrine subenvironments (Eberle and Lillegraven, 1998; Wroblewski, 2003, 2004). In the Williston Basin, the latest Cretaceous is represented by the Hell Creek Formation, a unit that consists of ~100 m of mud-stones interbedded with sandstones, interpreted as the rem-nants of ancient, meandering, aggradational fluvial systems (Fastovsky, 1987; Murphy et al., 2002).

Like some great cosmic joke designed to frustrate pa leontologists, the K-T boundary is rarely found in a thick sequence of strata unambiguously representing continuous deposition in a single paleoenvironmental setting. In the Hanna Basin, the K-T boundary probably occurs adjacent to or within a disconformity-bounded, 8-m-thick, complexly channeled sandstone, the so-called “zone of uncertainty” of Lillegraven and Eberle (1999). By contrast, in the Williston Basin, the boundary interval is characterized by a facies

change in which the interbedded mudstones and sandstones of the Hell Creek Formation generally give way to extensive laminated siltstones and coal deposits of the basal part of the Fort Union Formation (Fastovsky, 1987; Lofgren, 1995). The lithostratigraphic contact between the Hell Creek and Fort Union Formations is not precisely isosynchronous, but var-ies by as much as 3 m above or below the palynologically identified K-T boundary (see Fastovsky, 1987; Johnson, 1992; Lund et al., 2002; Nichols and Johnson, 2002; Pearson et al., 2002). The low density of dinosaur preservation (an estimated 0.000056 dinosaurs/m2 of exposure; White et al., 1998) means that in practice, the facies change is commonly (but certainly not always) associated with the dinosaur extinction.

RECONSTRUCTING FLUXES IN DINOSAUR DIVERSITY AT THE K-T BOUNDARY

Over the past 15 years, three studies, by three different research groups using different approaches, have produced stratigraphically refined, quantitative data bearing upon the rate of the Late Cretaceous dinosaur extinction in western North America.

Hanna BasinDinosaur distributions in the Hanna Basin were studied

by determining, with maximum refinement, the stratigraphic ranges of taxa preserved in the Ferris Formation. Stratigraphic relations were established using a combination of paly-nostratigraphy and North American Land Mammal “Ages” (Cifelli et al., 2004; Lofgren et al., 2004; see also Grimaldi et al., 2000). Seventy-six vertebrate-bearing localities contributed to the results of the study (Fig. 2). Regarding the pattern of dinosaur extinction, Lillegraven and Eberle (1999, p. 702) concluded “there exists little evidence for progressive reduc-tions in taxonomic diversity of the local dinosaurian fauna… Indeed, our collections show that, exclusive of the rare forms, most species of dinosaurs from the Ferris Formation are rep-resented up [to] … the eight meter thick zone of complexly channeled sandstone that is uncertain in age. …The late his-tory of local dinosaurs seemed to have a sudden termination. That is, high taxonomic diversity persisted late into the [latest Cretaceous]…”

The authors thus identified a geologically instantaneous extinction, but, given the issues of temporal resolution described above, noted that their study could not distinguish between events that took place on 1000 to 10,000 year time-scales and those that are much shorter.

Williston BasinSheehan et al. (1991) divided the Hell Creek into three

sequential stratigraphic windows to test whether community-level changes among dinosaur assemblages were compat-ible with long-term extinction scenarios. A methodologically controlled census of dinosaur remains through the thick-ness of the formation produced a minimum number of 556 in situ, precisely located individuals within the Hell Creek Formation. Using rarefaction to compare the communities among the three windows, Sheehan et al. (1991) found that the number, rank order, and relative proportions of families were unchanged, indicating stable communities through the formation, and concluded that “there is no statistically mean-ingful drop in the ecological diversity of dinosaurs through

Figure 1. The Western Interior of North America, showing in blue the Cretaceous-Tertiary Hanna, Williston, and contiguous Powder River (not discussed here) basins. Area shown in tan exhibits Laramide deformation; Laramide-age thrust faults shown as barbed lines. Redrawn from Hamilton (1988) and Lillegraven and Eberle (1999).

PowderRiverBasin

6 MARCH 2005, GSA TODAY

the Hell Creek. [O]ur data are compatible with abrupt extinc-tion scenarios” (p. 838). The Sheehan et al. (1991) study was criticized for its use of rarefaction by Hurlbert and Archibald (1995), who noted that the technique was “only weakly related to ... variables such as absolute numbers of dino-saur taxa, population densities, or extinction rates” (p. 881). Sheehan et al. (1996) countered that the critique was misdi-rected because these variables were not the ones under con-sideration. They noted that absence of a taxonomic decline was established without rarefaction (number and rank order of taxa were unchanged) and affirmed that the statistical treat-ment was only an effort to search for a community-level reor-ganization such as might indicate a deteriorating ecosystem.

Approaching the question from a biostratigraphic perspec-tive, Pearson et al. (2002) undertook a 10-year effort to deter-mine the stratigraphic distribution of taxa in the Hell Creek (Fig. 3). At each of the 82 sites distributed among 20 localities that they studied, they identified the K-T boundary via pol-len and then surveyed the vertebrate faunas found in each locality. Their database included 2233 dinosaur specimens

distributed among 14 taxa and located to precise stratigraphic horizon. The results showed no obvious change in diversity through the Hell Creek, an inference that they statistically tested via rarefaction. They concluded, “there is no evidence for a decreasing trend [in diversity] through the formation … these results are not compatible with gradual extinction at the end of the Cretaceous” (p. 164–165).

All published, quantitative, stratigraphically refined, field-based studies involving dinosaurs are sending the same clear signal of geologically instantaneous extinction.

THE RECOVERYOne may also approach this problem from a quite different

perspective: A catastrophic perturbation to the ecosystem, as has been proposed, ought to leave some kind of mark in the fossil record, reflected in the subsequent recovery of the eco-system. And indeed, it has been common knowledge almost since dinosaurs were first identified as Mesozoic beasts that the terrestrial K-T boundary is characterized by a stunning ecosystemic turnover marked by the extinction of dinosaurs and the radiation of mammals. The nature of these early mammals has also long been well known: they were small, presumably omnivorous, generalists (e.g., Lillegraven et al., 1979; Maas and Krause, 1994). While their connection to more recent mammals is becoming manifest (e.g., Archibald et al., 2001; Archibald, 2002, 2003), and while modern clades of pla-cental mammals may have roots prior to the K-T boundary,

Figure 2. Stratigraphic range chart showing the distribution of non-mammalian fossil vertebrate specimens in the Ferris Formation in the Hanna Basin (data from Lillegraven and Eberle, 1999). Dinosaurs essentially disappear at the base of the 8-m-thick, largely unfossiliferous, “zone of uncertainty” (see text). Three exceptions to this are an isolated tooth found within the “zone of uncertainty” and two water-worn specimens, interpreted by Lillegraven and Eberle (1999) as Cretaceous material reworked in earliest Tertiary time. Formation thickness and key lithological changes, biostratigraphic zonations (North American Land Mammal “Ages” and “Interval Zones”), and geochronology (Ma) are shown on left. Dates are ±0.1 m.y. (interpolated from Lofgren et al., 2004; basal date from Grimaldi et al., 2000).

Figure 3. Composite stratigraphic range chart showing the distribution of identifiable fossil vertebrate specimens in the Hell Creek Formation, which records the last 1.4 m.y. of the Cretaceous in the Williston Basin. In the case of dinosaurs, no decrease in diversity could be found. A single non-avian dinosaur in the basal part of the Fort Union Formation demonstrates the diachroneity of the lithostratigraphic contact between the Hell Creek and Fort Union, reaffirming that in places, the basal Fort Union is Cretaceous in age. Redrawn from Pearson et al. (2002).

GSA TODAY, MARCH 2005 7

there is little comparison to be made between the appearance of modern mammalian clades during Late Cretaceous time and the radiation that followed the K-T boundary.

Like the dinosaur extinction, mammalian evolution in the early Tertiary of North America has been evaluated quanti-tatively (Maas and Krause, 1994; Alroy, 1999; Archibald and Deutschman, 2001). All agree that earliest Tertiary mammals underwent high rates of speciation leading to a steep increase in rates of diversification during the first 5 m.y. of the Tertiary (Fig. 4). Indeed, seventeen of the eighteen orders of extant placental mammals did not exist before the K-T boundary (Archibald, 2002).

A range of evolutionary specializations took place during the early Tertiary, as mammals invaded ecospace abandoned by dinosaurs. As noted by Alroy (1998), “... the data are com-patible with the idea that the extinction of large terrestrial vertebrates such as dinosaurs at the K-T boundary opened up the larger end of the body size spectrum for occupation by mammals” (p. 733).

Just how quickly did the post–K-T mammalian radiation proceed? Eberle and Lillegraven (1998) record increasing body sizes among primitive eutherian mammals (condylarths) within 400,000 yr of the K-T boundary in the Hanna Basin. Still, Maas and Krause (1994) report that it took mammals 3–5 m.y. to develop the broad range of body sizes and diversity of ecologic specializations that have characterized all succes-sive, stable, mammalian faunas throughout the Cenozoic (e.g., Janis et al., 1998). An even longer estimate was provided by Kirchner and Weil (2000), who suggested that the recovery took as long as 10 m.y.

Rates of evolution were evidently fueled by multiple waves of immigration (Weil and Clemens, 1998; Lillegraven and Eberle, 1999; Clemens, 2002). Clemens (2002) summarizes

this viewpoint: “Recovery of the terrestrial vertebrate fauna [in eastern Montana] was not simply the product of an explosive evolutionary radiation of a few surviving stocks. Immigration from other areas played a major role in reconstitution of mammalian diversity” (p. 240).

Interestingly, all accounts suggest that the timing of the post–K-T radiation of mammals lags somewhat behind that observed by d’Hondt et al. (1996, 1998) for planktonic fora-minifera after the K-T boundary. However, these estimates of mammalian recovery intervals are in the range of those for other organisms following other mass extinction events (Jablonski, 1991).

The record of earliest Tertiary mammals, with its high rates of speciation and diversification and generally delayed onset of broad morphological specializations and size dispari-ties, appears very much like the kinds of faunas and floras that typically develop after ecological traumas: the so-called “disaster species” (Brenchley and Harper, 1998). Indeed, evo-lutionary patterns in the earliest Tertiary are generally termed a recovery (but see Lillegraven and Eberle, 1999), signaling that there was something from which to recover.

CAUSES: SO WHAT KILLED THE DINOSAURS?We have spent much of this article reviewing aspects of the

pattern (in this case, the rate) of the North American dinosaur extinction. Causal factors are necessarily much more difficult to identify, because while events can potentially be shown to be coincident, the demonstration of causality is far more problematical. Indeed, the extinction of the dinosaurs has a notorious, ongoing history of insouciant proposals unfettered by data (see Fastovsky and Weishampel, 2005).

Our interest, however, is in published models that are grounded in data and that are potentially testable. In this category, Archibald and Bryant (1990) and Archibald (1996, 1997), basing their interpretations upon patterns of vertebrate survivorship, proposed marine regression with associated habitat fragmentation as the ultimate cause of the K-T extinc-tions. In this model, multiplication and lengthening of river systems due to a marine regression led to a diminution and fragmentation of coastal plain habitats, in turn causing range reductions and eventual extinction. A variation on this model was proposed by Dingus and Rowe (1997), who suggested that the regression in combination with latest Cretaceous igneous activity and the asteroid impact caused the extinction of the dinosaurs.

The conclusion that the extinction of the dinosaurs was geologically instantaneous precludes longer-term causes (e.g., events on million to ten-million-year timescales). So, although survivorship patterns may be in accord with habitat fragmen-tation–based models, habitat fragmentation as the driving force for the dinosaur extinction is problematical, because it is linked in this case to a marine regression that occurred over a million or more years. Moreover, recent stratigraphic work summarized in Johnson et al. (2002) suggests that the Hell Creek was deposited rather quickly (over ~1.4 m.y.; Hicks et al., 1999, 2002) in a transgressive setting (the final transgression of the North American Western Interior Sea). This interpretation is concordant with a previously inferred rise in the water table (Fastovsky and McSweeney, 1987). The transgressive geological setting is antithetical to the proposed

Figure 4. Cenozoic mammalian ordinal diversity plotted as a function of time. Note the sharp rises in the diversity of condylarths at 65 Ma and the secondary evolutionary bursts of Plesiadapiformes (“Ples.”) and insectivores at 62–63 Ma. By 55 Ma, all of these orders of mammals started to become far less dominant as the larger, more diversified perissodactyls, artiodactyls, carnivorans, and rodents became important parts of successive global mammalian faunas. (From Alroy, 1999; also at http://www.nceas.ucsb.edu/~alroy/mammalorders.gif).

8 MARCH 2005, GSA TODAY

1To review the most pervasively cited example: Since amphibians are particularly sensitive to pH conditions, and they passed through the boundary largely unscathed, the asteroid impact could not have had a significant involvement in the extinction, because one of its predicted side affects—acid rain—would surely have affected amphibian survivorship (Weil, 1994; Archibald, 1996, 1997). In fact, a significant body of work suggests that the survival of amphibians is good evidence that deadly acid rain was not a major effect of the impact (see also d’Hondt et al., 1994; Maruoka and Koeberl, 2003; Retallack, 2004).

fluvial lengthening associated with the habitat fragmentation scenario and sug-gests that it was not likely a factor in the North American dinosaur extinction.

Death by AsteroidThe current “alternative hypothesis”

for the cause of the extinction of the dinosaurs is, of course, an asteroid impact with Earth. Schultz and d’Hondt (1996), using the morphology of the crater as an indicator of the angle and direction of the impact, proposed that the Western Interior of North America would bear the brunt of impact effects. In all scenarios, wholesale extinctions on extremely short timescales are pre-sumed to be a consequence of such an event. While the extinction cannot be shown to have occurred within hours, days, or weeks, extinction timescales can be constrained to a few tens of thousands of years or less. For this rea-son, what is known of the rate of the dinosaur extinction in North America is concordant with the predicted effects of an asteroid.

In this brave new world of an asteroid blight, how did some organisms survive and others became extinct? A number of workers (e.g., Weil, 1994; Archibald and Bryant, 1990; Archibald, 1996, 1997; Dingus and Rowe, 1997) have argued that the patterns of survivorship are not concordant with the expected effects of an asteroid impact. Such effects include global wildfires (Wolbach et al., 1988; Ivany and Salawich, 1993), acid rain (Prinn and Fegley, 1987; Zahnle, 1990; Retallack, 2004), and atmospheric dust (Alvarez et al., 1980).1

On the other hand, Sheehan and Hansen (1986) first proposed, for both marine and terrestrial faunas, that buffering against the effects of the K-T extinction might accrue from detritus feeding because animals in detritus-based food chains appear to have fared better than those dependent upon primary production. Sheehan and Fastovsky (1992) and Archibald (1993, 1996), using data from eastern Montana first presented in Archibald

and Bryant (1990), demonstrated that in the terrestrial realm, aquatic organisms fared better than their fully terrestrial counterparts. Sheehan and Fastovsky (1992) proposed that aquatic organ-isms were generally more likely than exclusively terrestrial organisms to be in detritus-based food chains (see also Retallack, 2004) and that among ter-restrial animals, those able to feed in detritus-based food chains (e.g., mam-mals) were more likely to survive than animals in food chains dependent on primary production (e.g., dinosaurs). In Sheehan and Fastovsky’s (1992) sce-nario, as in the original Alvarez et al. (1980) scenario, sunlight was blocked for an extended period of time, causing a collapse in primary production and holding out the hope of survivorship to only those who could live exclusively from detritus.

These ideas were recently revisited by Robertson et al. (2004), who quan-tified a previously proposed infrared thermal pulse from a global rain of hot spherules (splashed from the K-T impact), reassessed the patterns of biotic survivorship, and suggested that the short-term (hours-long), global pulse of intense infrared radiation was the primary killing agent. It would have caused severe thermal stress and ignited global wildfires that incinerated anything that could not shelter itself: “sheltering underground, within natural cavities, or in water was the fundamen-tal means to survival during the first few hours of the Cenozoic. Shelter was by itself not enough to guarantee survival, but lack of shelter would have been lethal” (Robertson et al., 2004, p. 760). This model differs from its antecedents because a heat pulse (and subsequent wildfires) rather than the cessation of photosynthesis, was the primarily kill-ing agent, and thermal sheltering, rather than detritus feeding, allowed survival.

We sincerely doubt that we have reported the last word on the cause of the extinction of the dinosaurs. This is in part because, as we have seen, the

number of studies reconstructing the pattern of the extinction is sparse and the temporal resolution of the data is limited. Moreover, in the case of the asteroid impact, although the enormous energy released by the impact made disastrous environmental consequences inevitable, a generally accepted model for all the environmental changes associated with the K-T impact has not yet emerged. Indeed, each of the potential effects of an asteroid impact has been the subject of considerable discussion. Among the putative effects to have undergone reconsideration are global wild fires (Belcher et al., 2003), acid rain (d’Hondt et al., 1994; Maruoka and Koeberl, 2003), and dark-ness, which may have been caused by aerosols rather than dust (Pope, 2002). Nonetheless, the extinction of the dino-saurs has been characterized by a rich intellectual debate as new methods of extracting highly refined stratigraphic data from sections are pioneered, as new interpretations of global events are entertained, and as an understanding of catastrophic events in earth history (e.g., Powell, 1998) is forged.

CONCLUSIONSIn the 25 years since Alvarez et al.

(1980) first proposed that an impact was responsible for the K-T extinc-tions, stratigraphic and paleoecologic evidence have come together to pre-sent a reasonably cohesive picture of a quick demise of the dinosaurs. Evidence from the rates of dinosaur extinction suggests that the extinction was geo-logically instantaneous; this conclusion in combination with the nature of the post-Cretaceous biologic recovery sug-gests that the extinction occurred on an extremely short, irresolvable timescale. While the exact killing mechanisms may or may not yet have been identified, all the data—including the rate of extinc-tion, the nature of the recovery, and the patterns of survivorship—are concor-dant with the hypothesis of extinction by asteroid impact.

GSA TODAY, MARCH 2005 9

ACKNOWLEDGMENTSGrateful thanks are due to J.D.

Archibald, D.A. Eberth, K. Howard, J.A. Lillegraven, G.J. Retallack, and A. Sweet, all of whom reviewed earlier drafts of this manuscript and contributed immensely to its improvement. However, the viewpoints expressed here are not necessarily those of the reviewers. This work was supported in part by the Instituto de Geología, Universidad Nacional Autónoma de México, and by a Fulbright Garcia-Robles Award.

REFERENCES CITEDAlroy, J., 1998, Cope’s rule and the dynamics of body mass evolution in North American mammals: Science, v. 280, p. 731–734, doi: 10.1126/science.280.5364.731.

Alroy, J., 1999, The fossil record of North American mammals: evidence for a Paleocene evolutionary radiation: Systematic Biology, v. 48, p. 107–118, doi: 10.1080/106351599260472.

Alvarez, L.W., Alvarez, W., Asaro, F., and Michel, H.V., 1980, Extraterrestrial cause for the Cretaceous-Tertiary extinction: Science, v. 208, p. 1095–1108.

Archibald, J.D., 1993, Major extinctions of land-dwelling vertebrates at the Cretaceous-Tertiary boundary, eastern Montana: Comment: Geology, v. 21, p. 90–93, doi: 10.1130/0091-7613(1993)0212.3.CO;2.

Archibald, J.D., 1996, Dinosaur Extinction and the End of an Era: What the Fossils Say: New York, Columbia University Press, 237 p.

Archibald, J.D., 1997, Extinction, Cretaceous, in Currie, P.J., and Padian, K., eds., Encyclopedia of dinosaurs: New York, Academic Press, p. 221–230.

Archibald, J.D., 2002, Fossil evidence for a Late Cretaceous origin of “hoofed” mammals: Science, v. 272, p. 1150–1153.

Archibald, J.D., 2003, Timing and biogeography of the eutherian radiation: fossils and molecules compared: Molecular Phylogeny and Evolution, v. 28, p. 350–359, doi: 10.1016/S1055-7903(03)00034-4.

Archibald, J.D., and Bryant, L., 1990, Differential Cretaceous-Tertiary extinctions of non-marine vertebrates: Evidence from northeastern Montana, in Sharpton, V.L. and Ward, P., eds., Global catastrophes in earth history: An in-terdisciplinary conference on impacts, volcanism, and mass mortality: Geological Society of America Special Paper 247, p. 549–562.

Archibald, J.D., and Deutschman, D., 2001, Quantitative analysis of the timing of origin of placental orders: Journal of Mammal Evolution, v. 8, p. 107–124, doi: 10.1023/A:1011317930838.

Archibald, J.D., Averianov, A.O., and Ekdale, E.G., 2001, Late Cretaceous relatives of rabbits, rodents, and other eutherian mammals: Nature, v. 414, p. 62–65, doi: 10.1038/35102048.

Bakker, R.T., 1986, The dinosaur heresies: New York, William Morrow and Company, 481 p.

Belcher, C.M., Collinson, M.E., Sweet, A.R., Hildebrand, A.R., and Scott, A.C., 2003, Fireball passes, nothing burns: Geology, v. 31, p. 1061–1064, doi: 10.1130/G19989.1.

Brenchley, P.J., and Harper, D.A.T., 1998, Palaeoecology: Ecosystems, environments and evolution: London, Chapman & Hall, 402 p.

Cifelli, R.L., Eberle, J.J., Lofgren, D.L., Lillegraven, J.A., and Clemens, W.A., 2004, Mammalian biochronology of the latest Cretaceous, in Woodburne, M.O., ed., Late Cretaceous and Cenozoic Mammals of North America: Biostratigraphy and Geochronology: New York, Columbia University Press, p. 21–41.

Clemens, W.A., 2002, Evolution of the mammalian fauna across the Cretaceous-Tertiary boundary in northeast-ern Montana and other areas of the Western Interior, in

Hartman, J., Johnson, K.R., and Nichols, D.J., eds., The Hell Creek Formation and the Cretaceous-Tertiary boundary in the northern Great Plains: Geological Society of America Special Paper 361, p. 217–245.

Clemens, W.A., Archibald, J.D., and Hickey, L.J., 1981, Out with a whimper not a bang: Paleobiology, v. 7, p. 293–298.

d’Hondt, S.L., Pilson, M.E.Q., Sigurdsson, H., Hanson, A.K., and Carey, S., 1994, Surface water acidification and extinc-tion at the Cretaceous-Tertiary boundary: Geology, v. 22, p. 983–986, doi: 10.1130/0091-7613(1994)0222.3.CO;2.

d’Hondt, S.L., Herbert, T.D., King, J., and Gibson, C., 1996, Planktic foraminifera, asteroids, and marine production: Death and recovery at the Cretaceous-Tertiary boundary, in Ryder, G., Fastovsky, D.E., and Gartner, S., eds., The Cretaceous-Tertiary Event and other catastrophes in earth history: Geological Society of America Special Paper 307, p. 303–318.

d’Hondt, S.L., Donaghay, P., Zachos, J.C., Luttenberg, D., and Lindinger, M., 1998, Organic carbon fluxes and ecolog-ical recovery from the Cretaceous/Tertiary mass extinction: Science, v. 282, p. 276–279, doi: 10.1126/science.282.5387.276.

Dingus, L., and Rowe, T., 1997, The mistaken extinction: New York, W.H. Freeman and Company, 332 p.

Dodson, P., 1990, Counting dinosaurs: Proceedings of the National Academy of Sciences of the United States of America, v. 87, p. 7608–7612.

Dodson, P., 1996, The horned dinosaurs: Princeton, New Jersey, Princeton University Press, 346 p.

Eberle, J.J., and Lillegraven, J.A., 1998, A new, important, record of earliest Cenozoic mammalian history: Geologic setting, Multituberculata, and Peradectia: Rocky Mountain Geology, v. 33, p. 3–47.

Fastovsky, D.E., 1987, Paleoenvironments of vertebrate-bear-ing strata during the Cretaceous-Paleogene transition in eastern Montana and western North Dakota: Palaios, v. 2, p. 282–295.

Fastovsky, D.E., and McSweeney, K., 1987, Paleosols span-ning the Cretaceous-Paleogene transition, eastern Montana and western North Dakota: Geological Society of America Bulletin, v. 99, p. 66–77.

Fastovsky, D.E., and Weishampel, D.B., 2005, The evolution and extinction of the dinosaurs (second edition): Cambridge, Cambridge University Press, 485 p.

Fastovsky, D.E., Huang, Y., Hsu, J., Martin-McNaughton, J., Sheehan, P.M., and Weishampel, D.B., 2004, The shape of Mesozoic dinosaur richness: Geology, v. 32, p. 877–880, doi: 10.1130/G20695.1.

Grimaldi, D.A., Lillegraven, J.A., Wampler, T.W., Bookwalter, D., and Shedrinsky, D., 2000, Amber from Upper Cretaceous through Paleocene strata of the Hanna Basin, Wyoming, with evidence for source and taphonomy of fossil resins: Rocky Mountain Geology, v. 35, p. 163–204.

Hamilton, W.B., 1988, Laramide crustal shortening, in Schmidt, C.J., and Perry, W.J. Jr., eds., Interaction of the Rocky Mountain foreland and the Cordilleran thrust belt: Geological Society of America Memoir 171, p. 27–39.

Hicks, J.F., Obradovich, J.D., and Tauxe, L., 1999, Magnetostratigraphy, isotopic age calibration, and inter-continental correlation of the Red Bird section of the Pierre Schale, Niobrara County, Wyoming, USA: Cretaceous Research, v. 20, p. 1–27, doi: 10.1006/cres.1998.0133.

Hicks, J.F., Johnson, K.R., Obradovich, J.D., Tauxe, L., and Clark, D., 2002, Magnetostratigraphy and geochronology of the Hell Creek and basal Fort Union formations of south-western North Dakota and a recalibration of the age of the Cretaceous-Tertiary boundary, in Hartman, J., Johnson, K.R., and Nichols, D.J., eds., The Hell Creek Formation and the Cretaceous-Tertiary Boundary in the northern Great Plains: Geological Society of America Special Paper 361, p. 35–55.

Hurlbert, S.H., and Archibald, J.D., 1995, No statistical support for sudden (or gradual) extinction of dinosaurs: Geology, v. 23, p. 881–884, doi: 10.1130/0091-7613(1995)0232.3.CO;2.

Ivany, L.C., and Salawich, P.J., 1993, Carbon isotopic evi-dence for biomass burning at the K-T boundary: Geology, v. 21, p. 487–490, doi: 10.1130/0091-7613(1993)0212.3.CO;2.

Jablonski, D., 1991, Extinctions: A paleontological perspec-tive: Science, v. 253, p. 754–757.

Janis, C.M., Scott, K.M., and Jacobs, L.L., editors, 1998, Evolution of Tertiary mammals in North America: Cambridge, Cambridge University Press, 703 p.

Johnson, K.R., 1992, Leaf-fossil evidence for extensive floral extinction at the Cretaceous-Tertiary boundary, North Dakota, USA: Cretaceous Research, v. 13, p. 91–117, doi: 10.1016/0195-6671(92)90029-P.

Johnson, K.R., Nichols, D.J., and Hartman, J.H., 2002, Hell Creek Formation: a 2001 synthesis, in Hartman, J., Johnson, K.R., and Nichols, D.J., eds., The Hell Creek Formation and the Cretaceous-Tertiary Boundary in the northern Great Plains: Geological Society of America Special Paper 361, p. 503–510.

Kirchner, J.W., and Weil, A., 2000, Delayed biotic recovery from extinctions throughout the fossil record: Nature, v. 404, p. 177–180, doi: 10.1038/35004564.

Lehman, T., 1987, Late Maastrichtian paleoenviron-ments and dinosaur biogeography in the western interior of North America: Palaeogeography, Palaeoclimatology, Palaeoecology, v. 60, p. 189–217, doi: 10.1016/0031-0182(87)90032-0.

Lillegraven, J.A., and Eberle, J.J., 1999, Vertebrate faunal changes through Lanican and Puercan time in southern Wyoming: Journal of Paleontology, v. 73, p. 691–710.

Lillegraven, J.A., Kielen-Jaworowska, S., and Clemens, W.A., editors, 1979, Mesozoic mammals: The first two-thirds of mammalian history: Berkeley, University of California Press, 311 p.

Lillegraven, J.A., Snoke, A.W., and McKenna, M.C., 2004, Tectonic and paleogeographic implications of late Laramide geologic history in the northeastern corner of Wyoming’s Hanna Basin: Rocky Mountain Geology, v. 39, p. 7–64.

Lofgren, D.L., 1995, The Bug Creek problem and the Cretaceous-Tertiary transition at McGuire Creek, Montana: University of California Publications in the Geological Sciences, v. 140, 185 p.

Lofgren, D.L., Lillegraven, J.A., Clemens, W.A., Gingerich, P.D., and Williamson, T.E., 2004, Paleocene biochronology: the Puercan through Clarkforkian Land Mammal Ages, in Woodburne, M.O., ed., Late Cretaceous and Cenozoic mammals of North America: Biostratigraphy and geochro-nology: New York, Columbia University Press, p. 43–105.

Lund, S.P., Hartman, J.H., and Banerjee, S.K., 2002, Magnetostratigraphy of interfingering upper Cretaceous-Paleocene marine and continental strata of the Williston Basin, North Dakota and Montana, in Hartman, J., Johnson, K.R., and Nichols, D.J., eds., The Hell Creek Formation and the Cretaceous-Tertiary Boundary in the northern Great Plains: Geological Society of America Special Paper 361, p. 57–74.

Maas, M.C., and Krause, D.W., 1994, Mammalian turn-over and community structure in the Paleocene of North America: Historical Biology, v. 8, p. 91–128.

Maruoka, T., and Koeberl, C., 2003, Acid-neutral-izing scenario after the Cretaceous-Tertiary impact event: Geology, v. 31, p. 489–492, doi: 10.1130/0091-7613(2003)0312.0.CO;2.

Murphy, E.C., Hoganson, J.W., and Johnson, K.R., 2002, Lithostratigraphy of the Hell Creek Formation in North Dakota, in Hartman, J., Johnson, K.R., and Nichols, D.J., eds., The Hell Creek Formation and the Cretaceous-Tertiary Boundary in the northern Great Plains: Geological Society of America Special Paper 361, p. 9–34.

Nichols, D.J., and Johnson, K.R., 2002, Palynology and microstratigraphy of Cretaceous-Tertiary boundary sections in southwestern North Dakota, in Hartman, J., Johnson, K.R., and Nichols, D.J., eds., The Hell Creek Formation and the Cretaceous-Tertiary Boundary in the northern Great Plains: Geological Society of America Special Paper 361, p. 95–143.

Pearson, D.A., Schaefer, T., Johnson, K.R., Nichols, D.J., and Hunter, J.P., 2002, Vertebrate biostratigraphy of the Hell Creek Formation in southwestern North Dakota and northwestern South Dakota, in Hartman, J., Johnson, K.R., and Nichols, D.J., eds., The Hell Creek Formation and the Cretaceous-Tertiary Boundary in the northern Great Plains: Geological Society of America Special Paper 361, p. 145–167.

Peterson, J.A., 1986, editor, Paleotectonics and sedimenta-tion in the Rocky Mountain region, United States: American Association of Petroleum Geologists Memoir 41, 693 p.

10 MARCH 2005, GSA TODAY

v. 24, p. 963–967, doi: 10.1130/0091-7613(1996)0242.3.CO;2.

Weil, A., 1994, K-T survivorship as a test of the acid rain hypothesis: Geological Society of America Abstracts with Programs, v. 26, no. 7, p. A-335.

Weil, A., and Clemens, W.A., 1998, Aliens in Montana: phy-logenetically and biogenetically diverse lineages contrib-uted to an earliest Cenozoic community: Geological Society of America Abstracts with Programs, v. 30, no. 5, p. 69–70.

Weishampel, D.B., Barrett, P.M., Coria, R.A., Le Loeuff, J., Xu, X., Zhao, X., Sahni, A., Gomani, E.M.P., and Noto, C.R., 2004, Dinosaur distribution, in Weishampel, D.B, Dodson, P., and Osmólska, H., eds., The Dinosauria, 2nd edition: Berkeley, University of California Press, p. 517–606.

White, P.D., Fastovsky, D.E., and Sheehan, P.M., 1998, Taphonomy and suggested structure of the dinosaurian assemblage of the Hell Creek Formation (Maastrichtian), eastern Montana and western North Dakota: Palaios, v. 13, p. 41–51.

Wolbach, W.L., Gilmour, I., Anders, E., Orth, C.J., and Brooks, R.R., 1988, A global fire at the Cretaceous/Tertiary boundary: Nature, v. 343, p. 152–154.

Wroblewski, A.F.-J., 2003, Tectonic redirection of Paleocene fluvial drainage systems and lacustrine flooding in the Hanna Basin area, south-central Wyoming, in Raynolds, R.G., and Flores, R.M., eds., Cenozoic systems of the Rocky Mountain region: Denver, Rocky Mountain Society for Sedimentary Geology (SEPM), p. 227–252.

Wroblewski, A.F.-J., 2004, New Selachian paleofaunas from “fluvial” deposits of the Ferris and lower Hanna formations (Maastrichtian-Selandian: 66–58 Ma), southern Wyoming: Palaios, v. 19, p. 249–258.

Zahnle, K., 1990, Atmospheric chemistry by large im-pacts: Geological Society of America Special Paper 247, p. 271–288.

Manuscript received October 11, 2004; accepted December 22, 2004.

Pope, K.O., 2002, Impact dust not the cause of the Cretaceous-Tertiary mass extinction: Geology, v. 30, p. 99–102, doi: 10.1130/0091-7613(2002)0302.0.CO;2.

Powell, J.L., 1998, Night comes to the Cretaceous: New York, W.H. Freeman and Company, 250 p.

Prinn, R.G., and Fegley, B., 1987, Bolide impacts, acid rain, and biospheric traumas at the Cretaceous-Tertiary boundary: Earth and Planetary Science Letters, v. 83, p. 1–15, doi: 10.1016/0012-821X(87)90046-X.

Retallack, G.J., 2004, End-Cretaceous acid rain as a selec-tive extinction mechanism between birds and dinosaurs, in Currie, P.J., et al., eds., Feathered dragons—Studies on the transition from dinosaurs to birds: Bloomington, Indiana University Press, p. 35–64.

Robertson, D.S., McKenna, M., Toon, O.B., Hope, S., and Lillegraven, J.A., 2004, Survival in the first hours of the Cenozoic: Geological Society of America Bulletin, v. 116, p. 760–768, doi: 10.1130/B25402.1.

Sheehan, P.M., and Fastovsky, D.E., 1992, Major extinc-tions of land-dwelling vertebrates at the Cretaceous-Tertiary boundary, eastern Montana: Geology, v. 20, p. 556–560, doi: 10.1130/0091-7613(1992)0202.3.CO;2.

Sheehan, P.M., and Hansen, T.A., 1986, Detritus feeding as a buffer to extinction at the end of the Cretaceous: Geology, v. 14, p. 868–870, doi: 10.1130/0091-7613(1986)142.0.CO;2.

Sheehan, P.M., Fastovsky, D.E., Hoffmann, R.G., Berghaus, C.B., and Gabriel, D.L., 1991, Sudden extinction of the di-nosaurs: Family-level patterns of ecological diversity during the latest Cretaceous, upper Great Plains, U.S.A.: Science, v. 254, p. 835–839.

Sheehan, P.M., Fastovsky, D.E., Hoffmann, R.G., and Barreto, C.B., 1996, Comment: No statistical support for sudden (or gradual) extinction of dinosaurs: Geology, v. 24, p. 957–958, doi: 10.1130/0091-7613(1996)0242.3.CO;2.

Schultz, P.H., and d’Hondt, S.L., 1996, Cretaceous-Tertiary (Chicxulub) impact angle and its consequences: Geology,