collection. Our data show, however, that 4. C. Koeberl and W. A. Cassidy, Lunar Planet. Inst.

ar Planet. Inst. model for initial calculations; our calculated

there is also a "background" source of Tech. Rep. 90-01 (1990).
5. J. E. Dennison and M. E. Lipschutz, Geochim. results are least accurate in the near field
<190'C group H5 chondrites (for exam- Cosmochim. Acta 51, 741 (1987). (within about 40 km of the rupture surface)
ple, Fig. 1), with cosmic-ray exposure ages 6. J. T. Wasson et al., ibid. 53, 735 (1989). where the details of the slip distribution are
7. J. T. Wasson, Science 249, 900 (1990).
of >20 Ma, which might represent earlier 8. E. Jarosewich, Lunar Planet. Inst. Tech. Rep. 90-01 important. In the far field, the details of the
breakup events or even fragments of a (1990), p. 54. slip distribution become less important than
completely different parent body. 9. H. Takeda, ibid., p. 86. the total moment of the earthquake and the
10. G. W. Wetherill, Meteoritics 20, 1 (1985).
In summary, we suggest that we can now 11. D. W. G. Sears, Nucl. Tracks Radiat. Meas. 14, 5 average orientation of the rupture plane.
account for the origin and destruction of a (1988). The major faults in the Bay Area were
large group of H5 chondrites found only in 12. M. Haq, F. A. Hasan, D. W. G. Sears, Geochim.
Cosmochim. Acta 52, 1679 (1988). represented by alignments ofvertical 10-km-
the Antarctic meteorite collection. Our data 13. K. Nishiizumi, D. Elmore, P. W. Kubik, Earth long rectangular patches extending from the
show that the terrestrial meteorite flux is not Planet. Sci. Lett. 93, 299 (1989). surface to a depth of 13 km (8). Stress
14. P. H. Benoit, H. Sears, D. W. G. Sears, J. Geophys.
a constant. Numbers, sizes, and relative pro- Res., in press. changes were calculated at the centers of the
portions of different meteorite types can 15. D. W. G. Sears, P. Benoit, J. D. Batchelor, Geochim. patches (9). Both shear stress and normal
change over relatively short periods of time, Cosmochim. Acta 55, 1193 (1991). stress changed on the fault surfaces as a
16. L. Schultz, H. W. Weber, F. Begemann, ibid., p. 59.
at least in some cases. In this light, we 17. J. A. Wood, Icarus 6, 1 (1967). result of the Loma Prieta earthquake. Be-
suggest that a more critical examination of 18. K. Nishiizumi, S. Regnier, K. Marti, Earth Planet. cause the major vertical Bay Area faults are
Sci. Lett. 50, 156 (1980).
the large Antarctic meteorite collection may probably loaded predominantly by right-
19. J. Willis and J. I. Goldstein, Proc. Lunar Planet. Sci.
well turn up more cases of temporal changes Conf 12, 1135 (1981). lateral shear, the shear stress changes im-
in the meteorite flux. 20. We thank the U.S. Antarctic Meteorite Working posed by the Loma Prieta rupture can either

Downloaded from www.sciencemag.org on December 24, 2010

Group and M. E. Lipschutz for samples and docu-
mentation, J. Roth and V. Yang for technical assist- increase (if the change is right-lateral) or
REFERENCES AND NOTES ance, and G. A. McKay for access to the Johnson decrease (if the change is left-lateral) the
Space Center electron microprobe. This research shear load. Laboratory studies of rock fric-
1. I. M. Whillans and W. A. Cassidy, Science 222, 55 was supported by National Aeronautics and Space
(1983). Administration grant NAG 9-81 and National Sci- tion and failure suggest that normal forces
2. J. T. Wasson, Meteorites: Their Record of Early ence Foundation grant DPP 86-13998. are important in determining resistance to
Solar-System History (Freeman, New York, 1985).
3. 0. Eugster, Science 245, 1197 (1989). 15 November 1991; accepted 31 January 1992 sliding (10). We defined a Coulomb failure
function (CFF) for comparison with
changes in rates of seismicity:

CFF IST-,u -p)-S

Response of Regional Seismicity to the Static Stress This function is based on the Coulomb
criterion for shear failure of rocks
Change Produced by the Loma Prieta Earthquake
ISTI > (gO - p) + S
PAUL A. REASENBERG AND ROBERT W. SIMPSON where I'rI is the magnitude of shear traction
acting on the plane, cr is the normal traction
The 1989 Loma Prieta, California, earthquake perturbed the static stress field over a (positive for compression), p is the pore
large area of central California. The pattern of stress changes on major faults in the fluid pressure, g. is the coefficient of internal
region predicted by models of the earthquake's dislocation agrees closely with changes friction, and S is the cohesion (11). Thus
in the regional seismicity rate after the earthquake. The agreement is best for models changes in the CFF are given by
with low values of the coefficient of friction (0.1 < p. c 0.3) on Bay Area faults. Both
the stress models and measurements suggest that stresses were increased on the San ACFF = 1TI - 'rTO - g(Ac- Ap) -AS
Andreas fault north ofthe Loma Prieta rupture, but decreased slightly on the Hayward This expression is nonlinear in the changes
fault. This relaxation does not warrant lower probability estimates for large earth- in shear stress (12). It can be simplified by
quakes on the Hayward fault in the next 30 years, however. assuming that for vertical strike-slip faults
the horizontal shear stresses are most impor-
T HE MAGNITUDE (M) 7.1 LoMA PRi- side (2). Because the Hayward fault proba- tant and that changes are superposed on a
eta earthquake was the largest earth- bly ruptured on both of these occasions, the preexisting background level of right-lateral
quake to strike the San Francisco Bay stress changes on the Hayward fault were of shear for most faults in the Bay Area. This
region since 1906 (1). In addition to radi- particular interest (3). To evaluate the pos- leads to
ating seismic waves, the earthquake intro- sibility of such interactions we determined ACFF A-rTih
duced a static stress perturbation to the the static stress changes produced within an g.Ao
region as a result of its displacement, which elastic half space by model dislocation sur- where AThrl is the change in horizontal com-
averaged about 2 m between depths of 6 and faces, and compared the results to regionally ponent of right-lateral shear stress (positive
18 km. An immediate concern after the observed changes in the rates of seismicity for more right-lateral) and Au is the change
earthquake was the possible effect this stress (4, 5). in normal stress (positive for more compres-
perturbation might have had on other major For our stress calculations we simulated sion) (13). We also assume here that changes
faults in the Bay Area. On two occasions in the Loma Prieta displacement at depth by a in p and S are negligible.
the 1800s a large earthquake on one side of rectangular dislocation surface inferred from As discussed below, a choice of = 0.2
the Bay had been followed within 3 years by geodetic measurements made before and yielded the best agreement with observed
a second large earthquake on the opposite after the earthquake (6). Although complex- seismicity rate changes (Fig. 1). Changes in
ities in the seismic waveforms observed dur- CFF ranged from a few bars to less than
U.S. Geological Survey, Mail Stop 977, 345 Middlefield ing the earthquake suggest that the rupture 0.01 bar (Fig. 1). San Andreas fault seg-
Road, Memo Park, CA 94025. was complex (7), we chose to use a simple ments at either end of the Loma Prieta

27 MARCH 1992 REPORTS 1687

Fig. 1. Static stress changes produced
by the Loma Prieta earthquake on
vertical planar segments representing
faults in central California, c ated a
using an elastic model with a rectan- -,

gular dislocation (yellow) derived

from geodetic measurements. Change
in the Coulomb failure fiuction
ACFF (in bars) on each fault segment
for ji = 0.2 is indicated by color. OR-

Segments with increased CFF (red)

are more likely to produce earth-
quakes; segments with decreased max,
CFF (blue), less likely. This oblique km
view is from the southwest and down
at an angle of 45°. Inset map shows
study area. Segment numbers are
shown for reference in Fig. 3. -5 -4 -3 -2 -1 0 1 2 3 4 5

rupture are closer to failure than before the earthquakes occurring in the respective inter- shocks ofmoderate earthquakes in the region
rupture because of the additional right-lateral vals. The rate change is expressed as the pre-Loma Prieta background period
in

Downloaded from www.sciencemag.org on December 24, 2010

stress loaded onto them. Estimates from sim- initially were detected with the complete cat-
ilar models of the time by which these stress na- E(ja) alog. To reduce these effects, we removed
changes might advance the occurrence of the (na, nb, ta) tb)
=
aftershocks from the catalog with a computer
next earthquakes on segments to the north of algorithm (21, 22).
the Loma Prieta rupture range from 2 to 25 where var denotes variance and E(na) = rbta is Areas of high postseismic seismicity rate
years (3). On the Hayward fault, changes in the value of na expected under the null hy- include the immediate aftershock zone
CFF range from 0.03 bar at the north end to pothesis of stationary random occurrence (from Los Gatos to Watsonville); the San
0.8 bar at the south end, in a sense to delay (16). We calculated 1P for fixed ta and tb in Francisco Peninsula in an area of thrust
the onset of the next earthquake. For a long- overlapping 10-km-square cells covering an faults near Los Gatos and along the San
term slip rate of 9 mm/year these changes in area 140 by 390 km for the preseismic (back- Andreas and San Gregorio faults near Daly
CFF amount to delays ranging from less than ground) period between 17 October 1979 City; along the San Andreas and Sargent
1 to 10 years, although these static, purely and 17 October 1989 and postseismic period faults from Watsonville to Bear Valley; a
elastic models may poorly estimate the longer between 18 October 1989 and 31 May 1991 partially offshore segment of the San Grego-
term effect (14). (17-19) (Fig. 2). Positive values of P indicate rio fault near Point Ano Nuevo; and along
In order to see if any of these stress changes that the postseismic rate was higher than the the Alamo fault, near Livermore. With the
were reflected in seismicity, we measured the background rate; negative values, lower (20). exception of the activity near Point Ano
regional coseismic change in seismicity rate A symmetry in the definition of introduces Nuevo (23), these areas generally coincide
using the statistic A, which is sensitive to a an ambiguity in interpretation, whereby a with fault segments having increased CFF.
contrast of average seismicity rate between relatively low postseismic rate cannot be dis- Areas of apparent coseismic decrease in seis-
two time intervals in a specified area (15). We tinguished from an abnormally high preseis- micity rate occur along the southern Cala-
compared the rate r4 in the postseismic inter- mic rate, as would result, for example, from veras, Hayward, southern Rodger's Creek,
val of duration ta with the rate rb in the an earthquake swarm or aftershock sequence and Mount Lewis faults; these fault seg-
preseismic interval of duration tb, where r4 = in the preseismic (background) period. Sev- ments also have relaxed stress (ACFF < 0).
nla/ta, rb = nfltb, and n,, nb are the numbers of eral such artifacts corresponding to the after- Other aftershock sequences before 1989

Fig. 2. Changes in seismicity rate

in central California occurring at
the time of the Loma Prieta earth-
quake. Values of (color) compare
average rates in the intervals 17
October 1979 to 17 October 1989
and 18 October 1989 to 30 June
1991. Areas that experienced a
coseismic increase (decrease) in
seismicity rate are shown as red
(blue); areas with insufficient earth-
quakes for determination of are
white. Absolute value of indicates 1w
the statistical significance ofthe rate 0 50 100 150 200 KM
change, taking into account the .~~~~~~~~~---

variance ofthe background seismic-

ity. Major faults and coastline are
indicated by solid lines. -5 4 -3 -2 -1 0 1 2 3 4 5
1688 SCIENCE, VOL. 255
Fig. 3. Seismicity change P and flow anomaly associated with the San An-
stress change ACFF(Ii) on 141 dreas fault (32). The apparent low friction
model fault segments for g. = 0.2. 25- value could also be explained by a high
Numbers (refer to Fig. 1) indicate
segments experiencing the largest coefficient of friction (consistent with labo-
changes (1oll 2 and IACFFI a 20O4 a ratory studies) combined with changes in
0.2 bar). pore pressure in a compliant fault zone (33).
Our observations support the conclusion
(3) (based on stress models alone) that sec-
tions of the San Andreas fault north of the
0.~~~~~~~~~~~~~~ Loma Prieta earthquake were probably
moved closer to failure as a result ofthe stress
change produced by the earthquake. In addi-
-0~~~
-2- 0m
Chang inCuobfiuefnto br tion, the Hayward fault may have relaxed
slightly. Such relaxation may be only tempo-
rary, however. Simple models suggest that
4 nonelastic adjustments and continued loading
with time may alter both the magnitude and
sign of stress changes induced on other faults
by the Loma Prieta earthquake (34). The

Downloaded from www.sciencemag.org on December 24, 2010

12 occurred on 87 fault segments, discor'dant minimal degree of relaxation, uncertainties in
11
1
1
1
1
1
changes occurred on 54 segments, a! nd a its estimation, and its temporary effect offer
10 1 generally positive correspondence bets ween no basis for reducing the 30-year probabilities
11
1
8 11
1
in.
g and ACFF is visually apparent. A x' test for large earthquakes in central California or
1
1 - .\-.k
1
on all fault segments rejects the null hyploth- the regional efforts to prepare for them.
0
E6. esis that g and ACFF(IL) are independernt (p Our result suggests that the Loma Prieta
41
<0.001 for 0.2 < [L < .3;p <0.01 fo r 0.1 earthquake will not trigger an earthquake on
* ~ ~ ~ ~ ~~~~ u
< Ii S 0.4) (25). The correlation coeffi cient the Hayward or Mission fault. This conclu-
0.0 0.2 0.4 0.6 p attains its maximum value for 1. = 0.2,, and sion, however, relies on a simple structural
Coefficient of friction exceeds 0.5 for 0.1 < g _ 0.3 (Fig. 4) (26). model consisting ofvertical faults. The pres-
Fig. 4. Measures of agreement between stress Static stress changes as small as a few teenths ence of dipping structures along the Hay-
changes and seismicity changes on fault segments, of a bar apparently produced detectable ward fault could change this result (35).
shown as a function ofthe coefficient of friction g. changes in seismicity. This level of sstress Some of our models with nonvertical Hay-
Squares, values of x2 (scale on left); triangles, change is about one order of magniitude ward and Mission faults weakly support a
coefficients of correlation between stress change
and seismicity change (scale on right). Solid lines, larger than that of tidally induced sstress triggering hypothesis, but these are poorly
all 141 fault segments; broken lines, fault seg- changes (27), and is comparable to sstress constrained by or inconsistent with the seis-
ments experiencing significantly large changes changes produced at seismogenic depth by is micity (36). If such a pairing does occur,
(numbered points in Fig. 3). Chi-squared confi- the filling of water reservoirs (28). The re- however, it would suggest that our present
dence levels are 6.64 (p = 0.01) and 10.83 (p = structural model is inadequate or that non-
sponse of seismicity to stress change decreases
0.001).
with distance from the dislocation, anid is elastic effects have exceeded the static elastic
statistically undetectable (at the p = 0.05 stress changes, or both.
preclude unambiguous interpretation of the level) at distances beyond 80 to 100 kr n; at
negative values of Pi observed along the this distance the maximum absolute vali ue of REFERENCES AND NOTES
southern Calaveras and Mount Lewis faults ACFF is approximately 0.1 bar (29). The
1. U.S. Geological Survey Staff, Science 247, 286
in terms of postseismic effects. The Hayward apparent sensitivity of regional seismicity to (1990).
fault, however, was essentially free of signif- such small stress changes suggests that flu uctu- 2. The 1836 (M - 6.8) earthquake on the northern
icant swarm and aftershock activity during ations in regional seismicity may be usecd to Hayward fault was followed in 1838 by an (M ~

7.0) earthquake on the San Francisco Peninsula

the 1980s. Consequently, the apparent low detect and model some aseismic slip evrents, section of the San Andreas fault, and the 1865 (M -
postseismic rate observed there (strongest including afterslip, slow earthquakes, sliip on 6.5) earthquake, possibly on the Loma Prieta seg-
on the southern Hayward fault) is not be- the ductile portion of vertical faults, andA slip ment of the San Andreas fault, was followed in 1868
by an (M ~
7.0) earthquake on the (possibly
lieved to be an artifact and may be related to on horizontal detachment surfaces. southern) Hayward fault [W. L. Ellsworth, U.S.
the Loma Prieta earthquake. Our observations of the regional seisi;mic- Geol. Surv. Prof. Pap. 1515 (1990), p.153].
ity response to the Loma Prieta earthquake 3. Working Group on California Earthquake Probabil-
To investigate the overall dependence be- ities, U.S. Geol. Surv. Circ. 1053 (1990). The work-
tween the modeled stress changes and the favor models involving low friction (0.1 < FL ing group considered in detail the effect of an
observed seismicity, we compared, for each < 0.3) on central California faults. In con- increase in right-lateral shear stress on segments of
the San Andreas fault north of the Loma Prieta
fault segment S,, the mean seismicity rate trast, laboratory experiments on frictional earthquake rupture. They concluded that the effect
change index R. corresponding to the subset slip in rock that typically have indiccated was to shorten the time to the next earthquake on
of cells within 5 km of segment S. to the higher coefficients of friction (0.5 c g < these segments.
4. Other studies have suggested that changes in static
coseismic static stress change ACFF,(1i) cal- 0.8) have been widely used as an analo g for stress might affect patterns or focal mechanisms of
culated on that segment, for various as- brittle faulting in the upper 15 to 20 k of m aftershocks or trigger other earthquakes. These in-
dude R. S. Stein and M. Lisowski, J. Geophys. Rev.
sumed values of the coefficient of friction g the crust (30). Our result is consistent with 88, 6477 (1983); G. M. Mavko, S. Schultz, B. D.
(24). We show one of the better fitting the idea that low friction could explain Brown, Bull. Seirmol. Soc. Am. 75, 475 (1985); L.
comparisons (for = 0.2) in Fig. 3. For this seismological and other evidence for ffault- Erickson (5); D. H. Oppenheimer et al. (12); C. M.
Poley, A. G. Lindh, W. H. Bakun, S. S. Schulz,
value of concordant changes in stress and
p, normal compression in central Califtornia Nature 327,134 (1987); K. W. Hudnut, L. Seeber,
seismicity (both increase or both decrease) (12, 31) and the lack of an observable heat J. Pacheco, Geophys. Res. Lett. 16, 199 (1989); P.
27 MARCH 1992 REPORTS 1689
 A. Rydelek and I. S. Sacks, Geophys. J. Int. 100, 39 Removal of aftershocks reduced (but did not com- 0.5 for 0.0 < p < 0.3.
(1990); L. Seeber and J. G. Armbruster, Geophys. pletely eliminate) these artifacts. The low apparent 27. T. H. Heaton, Bull. Seismol. Soc. Am. 72, 2181
Res. Lett. 17, 1425 (1990); A. J. Michael, W. L. postseismic rate (-3 ' 1 < 0) remaining in some of (1982); M. McNutt and T. H. Heaton, Caltf Geol.
Ellsworth, D. 0. Oppenheimer, ibid., p. 1441; A. J. these areas (for example, along the Calaveras and 34, 12 (1981); F. D. Stacey, Physics of the Earth
Michael, J. Geophys. Res. 96, 6303 (1991). Mount Lewis faults) still hampers interpretation of (Wiley, ed. 2, New York, 1977).
5. L. Erickson, thesis, Stanford University (1986). these features as postseismic effects. 28. E. A. Roeloffs, J. Geophys. Res. 93, 2107 (1988).
6. M. Lisowski, W. H. Prescott, J. C. Savage, M. J. 22. Apparent changes in seismicity rate may result from 29. Using g = 0.2 we calculated x2 for subsets of
Johnston, Geophys. Res. Lett. 17, 1437 (1990). The man-made factors, including changes in network segments in various ranges of distance from the
model dislocation strikes at N44W, dips 70SW, operations and method used to calculate earthquake center of the earthquake dislocation. x2 is maxi-
extends 37 km horizontally, has its top at a depth of magnitude. R. E. Habermann and M. S. Craig mum (X2 > 21) for the 80 segments between 0 and
5 km and its bottom at a depth of 17.5 km. [Bull. Seismol. Soc. Am. 78, 1225 (1988)] suggest- 100 km from the source; X2 just fails to exceed the
Right-lateral slip of 1.66 m and reverse slip of 1.19 ed the presence of such shifts in the catalog during p = 0.05 critical point for the farthest 70 segments
m yield a moment of 3.0 x 1019 N - m for this the 1970s. However, because the period 1984 to located 80 or more km from the earthquake.
model. More complex models of the rupture were 1991 is believed to be free of such effects [D. 30. See, for example, J. D. Byerlee and W. F. Brace, J.
also used to predict the static stress changes. Al- Oppenheimer, personal communication], we do not Geophys. Res. 73, 6031 (1968) and J. H. Dieterich,
though the results with these models vary in their believe that the seismicity changes we observed are in Mechanical Behavior of Crustal Rocks, N. Carter,
details, our overall result is apparently not sensitive significantly affected by such artifacts. M. Friedman, J. Logan, D. Stearns, Eds. (American
to the choice of rupture model. 23. Segments 129, 130, and 131 experienced sizable Geophysical Union, Washington, DC, 1981), pp.
7. See, for example, G. Beroza, Bull. Seismol. Soc. Am. seismicity rate increases and calculated decreases in 103-120.
81, 1603 (1991). CFF(p.) for all values of friction (see Fig. 3). How- 31. M. D. Zoback et al., Science 238, 1105 (1987), and
8. In addition, two nonvertical fault segments (198 ever, as illustrated by the dipping segment (200) references therein.
and 199) were included, representing dipping thrust adjacent to segment 130, thrust mechanisms would 32. A. H. Lachenbruch and A. McGarr, U.S. Geol.
planes northeast of the San Andreas fault near Los be favored in this vicinity. Furthermore, earthquake Surv. Prof Pap. 1515, 261 (1990).
Gatos. A third dipping plane (200) adjacent to the focal mechanism solutions suggest that a majority of 33. For discussions of pore pressure effects see, for
San Gregorio fault was not used in the statistical the earthquakes there were oblique right-lateral example, J. Byerlee, Geophys. Res. Lett. 17, 2109
evaluation, but was added for illustrative purposes thrust events on dipping planes. (1990); J. R. Rice, in Fault Mechanics and Transport
[see (23)]. 24. We used values of 1 obtained with aftershocks Properties ofRock, B. Evans and T.-F. Wong, Eds.

Downloaded from www.sciencemag.org on December 24, 2010

9. The depth of the segment centers (6.5 km) coincides removed for the background period 1979 to 1989. (Academic Press, London, in press).
with the median depth of seismicity used in this 25. Evaluation of a fourfold table with a two-sided x2 34. M. F. Linker and J. R. Rice, Eos 72, 310 (1991).
study. test was carried out on 141 segments. Significance 35. The presence of a buried thrust under the Mission
10. J. D. Byerlee, Pure Appl. Geophys. 116,615 (1978). levels for x2 are 3.84 (p = 0.05), 6.64 (p = 0.01) fault was suggested by D. J. Andrews, Eos 72, 446
11. J. C. Jaeger and N. G. W. Cook, Fundamentals of and 10.83 (p = 0.001) [see, for example, L. Sachs, (1991). Also see R. S. Stein and G. C. P. King,
Rock Mechanics (Chapman and Hall, ed. 3, London, Applied Statistics (Springer-Verlag, New York, Science 224, 869 (1984); D. L. Jones, A. J. H.
1979). 1982), pp. 346-351]. Lomax, T. V. McEvilly, Eos 72, 446 (1991).
12. D. H. Oppenheimer, P. A. Reasenberg, R. W. 26. Many of the points in Fig. 3 indicate absolute 36. W. L. Ellsworth, J. A. Olson, L. N. Shijo, S. M.
Simpson, J. Geophys. Res. 93, 9007 (1988). changes in stress or seismicity on individual seg- Marks, Calif: Div. Mines Geol., Spec. Publ. 62
13. For the two dipping fault segments (198 and 199) ments comparable to the uncertainties associated (1982).
representing reverse faults, ACFF was calculated for with modeling errors (stress) and stochastic vari- 37. We thank R. Archuleta, W. Ellsworth, A. Lindh, M.
the assumption that the dominant shear stress on ance (seismicity). When these tests were applied Linker, A. Michael, D. Oppenheimer, J. Rice, and
these planes before the Loma Prieta earthquake was to a subset of fault segments with significant P. Segall for helpful discussion and comments and
oriented up-dip in a thrust sense. changes in stress and seismicity (numbered points thoughtful reviews of the manuscript.
14. The ranges in these estimates reflect the range in in Fig. 3), similar results were obtained. x2: p
distance of the segments from the Loma Prieta < 0.01 for 0.1 s g s 0.4; correlation: p exceeds 22 November 1991; accepted 3 February 1992
earthquake and uncertainty about rates of loading
and magnitudes of earlier earthquakes on these fault
segments. No evidence currently justifies estimating
the time of advance or delay of the next large
earthquake from the modeled average static stress
change on a fault. The earthquake process and
crustal structure are complex, and the existence of a The Fossil Record and Evolution: Comparing Cladistic
simple relation between stress change and time to
the next large earthquake is as yet unproved. and Paleontologic Evidence for Vertebrate History
15. M. V. Matthews and P. A. Reasenberg, Pure Appl.
Geophys. 126, 357 (1988). The rate change index X
is defined with respect to the null hypothesis, which
MARK A. NORELL AND MICHAEL J. NOVACEK
states that the earthquake occurrence has a Poisso-
nian distribution with the postseismic rate equal to The fossil record offers the only direct evidence of extinct life and thus has figured
the background rate.
16. The variance was represented by that of a binomial prominently in considerations of evolutionary patterns. But the incomplete nature of
process: var(n.) = nbt,. the fossil record has also been emphasized in arguments that fossils play only a
17. We used M 2 1.5 earthquakes in the U.S. Geolog-
ical Survey central California catalog.
secondary role in the recovery of phylogenetic histories based on extant taxa. Although
18. The period 1979 to 1989 produced an unusually these criticisms recently have been countered, there is no general understanding of the
large number of (M > 5) earthquakes in central correspondence between the fossil record and phylogeny. An empirical survey of
California, including the 1979 Coyote Lake (M = recently published studies suggests no basis for assuming that the stratigraphic
5.9),1980 Livermore (M = 5.9),1983 Coalinga (M occurrence of fossils always provides a precise reflection of phylogeny. Nevertheless,
= 6.7), 1984 Morgan Hill (M = 6.2), 1985 Ket-
tleman Hills (M = 5.5), 1986 Mount Lewis (M = our survey of a sample of taxa shows a tendency for positive correlation between age
5.7),1986 Quien Sabe (M = 5.7),1986 Alum Rock and dade rank and, hence, a degree of correspondence between phylogenetic pattern
(M = 5.3), and the 1988 and 1989 Lake Elsman (M
= 5.0, M = 5.2) earthquakes. This series of earth- and the paleontologic record.
quakes has been interpreted as possible evidence of
an intermediate-term precursory process leading to SINCE THE BIRTH OF PALEONTOLOGY specializations (1-3). However, what pat-
the Loma Prieta earthquake [L. R. Sykes and S. C.
Jaume, Nature 348, 595 (1990)]. the fossil record has been interpreted terns do this record reflect and how precisely
19. The background periods 17 October 1969 to 17 as a record of life's history. The pale- do they capture evolutionary events? For
October 1979 and 17 October 1969 to 17 October ontologic record of horses, for example, has instance, fossils have been considered to
1989 were also tried, and similar results were ob-
tained. Because of improved stability ofthe network, been claimed to demonstrate the potential of provide so little evidence for relationships
the background period 1979 to 1989 is considered fossils in disclosing the branching sequence among living taxa that it has been suggested
most reliable.
20. Significance levels for 111 estimated from its asymp- of taxa through time as well as indicating that they be relegated to a secondary role in
totic (Gaussian) distribution are 1.96 (p = 0.05) major evolutionary trends toward increasing reconstruction of phylogeny (4-6). Such
and 2.57 (p = 0.01). criticisms and recommendations are coun-
21. P. A. Reasenberg, J. Geophys. Res. 90, 5479
(1985). Results presented in this paper are based on Department of Vertebrate Paleontology, American Mu- tered in a recent demonstration that fossil
analysis of the catalog with aftershocks removed. seum of Natural History, New York, NY 10024. taxa preserve pivotal evidence for ancient
1690 SCIENCE, VOL. 255