How is the Parkfield Earthquake Experi-ment like technology stocks? They bothseemed like a great bet at the time. As

the first attempt to track a geological fault to failure, Parkfield (Fig. 1) is without peeranywhere in the world. This 25-km-longstretch of the San Andreas fault has hosted five (possibly six) shocks since 1857, all reach-ing magnitude 6 on the Richter scale. Not onlywere these earthquakes roughly the same insize and location, but they were also separatedby about the same amount of time, with successive intervals between earthquakes of24, 20, 21, 12 and 32 years. This pattern of roughly periodic events is bolstered by nearlyidentical seismograms for the three mostrecent earthquakes. In addition, the last twoshocks, in 1934 and 1966, were both precededby a magnitude-5 foreshock some 17 minutesbefore the main shock, several kilometres to the north.

Parkfield’s behaviour anchored the con-cept of time-predictable earthquake recur-rence, in which an earthquake strikes when a fault recovers the stress relieved by the pre-vious event1. This means that the larger theearthquake on a fault, the longer the waituntil the next one. With each Parkfield quakesimilar in size, and a constant loading rate,the next would occur in 22 years — in 1988 —with an uncertainty of about 10 years. And so, in 1986 the United States Geological Survey inaugurated a focused experiment tomeasure the strain accumulation, capture thenucleation of the next rupture, and watch itpropagate2. But the earthquake never arrived.

Even though the timing of the Parkfieldearthquakes has turned out to be more irreg-ular than envisaged, if the stress accumulatedsince 1966 has not exceeded the stressreleased in 1966, the earthquakes would stillbe time-predictable. But on page 287 of thisissue, Murray and Segall3 remove this lastrefuge. Drawing on 40 years of geodeticmeasurements, they rigorously estimate the‘seismic moment’ of the 1966 event (ameasure of the earthquake’s size — a productof the fault slip, the area of slip and the elasticstiffness of the crust), and compare it to the accumulating ‘moment deficit’ (themoment associated with slip that has notoccurred since the 1966 quake). Astonish-ingly, they find that the moment released in1966 had been fully recovered by 1987. Theexpected earthquake is thus at least 15 yearsoverdue, approaching a complete cycle.

So have Murray and Segall dethronedtime-predictability? The authors acknowl-edge a key caveat: faults are not stressed inisolation. Other earthquakes change thestress acting on the San Andreas fault, whichcould delay or advance the next Parkfieldevent. Candidates include the 1983 Coalingaearthquake, 25 km to the east and magnitude6.5, which might have delayed the Parkfieldearthquake by a decade4; and a 1993 event5,magnitude 4.6 and a few kilometres beneaththe focus of the 1966 quake, which mighthave advanced the next Parkfield earth-quake. Although the effects of such stresschanges might explain the delay, they cannotrescue time-predictability, because Murrayand Segall measure the full moment deficitbetween 1966 and 1998, which includes theeffects of the perturbing shocks. Thus, themoment imbalance remains inescapable.

But does a lone, overdue earthquake

invalidate an earthquake-recurrence hypoth-esis? Its proponents would argue that it is based on rational physical principles, and that it works better than the Poissonhypothesis, which assumes the timings ofearthquakes are random. Further, time-predictability, or at least periodic earthquakeoccurrence, has been found to occur for somevery small ‘repeating’ shocks at Parkfield and elsewhere6. When applied to larger earthquakes, time-predictability works onlycrudely, and is thus applied with the explicitassumption of large expected variability.What is now needed are tests of time-pre-dictability on other faults. The ingredients ofa good test are fast fault-slip rates, moderate-sized earthquakes with short inter-eventtimes, and dense geodetic coverage. All of thispoints to Japan, where the concept was born.

While the Parkfield experiment can nolonger serve as a confirmation of time-

NATURE | VOL 419 | 19 SEPTEMBER 2002 | www.nature.com/nature 257

news and views

Parkfield’s unfulfilled promiseRoss S. Stein

Figure 1 The San Andreas fault near Cholame, California. Cholame lies at the southern end of theParkfield region, which has suffered five or six magnitude-6 earthquakes at regular intervals since 1857.

MIC

HA

EL

CO

LLIE

R

The idea that earthquakes are ‘time-predictable’ underlies many of today’sprobabilistic forecasts. In a key test on California’s San Andreas fault theconcept is found wanting, but the news may not be all bad.

© 2002 Nature Publishing Group

predictability, it is very much alive as a test ofslip-predictability, in which the longer thewait, the larger the next earthquake1. Murrayand Segall argue that an earthquake of mag-nitude 6.6–6.9 would balance the momentdeficit that has accumulated since 1966, andthe magnitude increases with each passingyear7. If, when the next Parkfield shockstrikes, its magnitude approaches this expec-tation, the Parkfield experiment would betransformed from a lesson in patience to aprescient success — tracking the nucleationand propagation of a much larger shock withParkfield’s arsenal of instruments would be a bigger scientific prize. And the observingpower at Parkfield has just taken a new leap,with the completion of a 2.2-km-deep pilotwell with a string of down-hole seismom-

eters, and the proposed 4.0-km borehole thatwill pierce the fault at seismogenic depths8.Perhaps in the end the delay will appear prov-idential, and Parkfield will turn out to be aninspired long-term investment for science. ■

Ross S. Stein is at the United States GeologicalSurvey, Menlo Park, California 94025, USA. e-mail: rstein@usgs.gov

1. Shimazaki, K. & Nakata, T. Geophys. Res. Lett. 7, 279–282 (1980).

2. Roeloffs, E. Curr. Sci. 79, 1226–1236 (2000).

3. Murray, J. & Segall, P. Nature 419, 287–291 (2002).

4. Toda, S. & Stein, R. S. J. Geophys. Res. 107,

10.1029/2001JB000172 (2002).

5. Fletcher, J. B. & Guatteri, M. Geophys. Res. Lett. 26, 2295–2298

(1999).

6. Nadeau, R. M. & McEvilly, T. V. Science 285, 718–721 (1999).

7. Harris, R. A. & Archuleta, R. J. Geophys. Res. Lett. 15,

1215–1218 (1988).

8. Thurber, C. et al. EOS Trans. Am. Geophys. Union Suppl. 81,

S318 (2000); www.earthscope.org/safod.html

occur with relative ease in juvenile birds, it was thought that adult owls were less malleable in this respect (Fig. 1). The sameprisms placed on an older owl, even for aperiod of months, cause only a tiny shift ofthe auditory map. In line with the terminol-ogy used in other kinds of neural systems, this window for large-scale auditory–visualrealignment might be called a ‘critical period’in development.

Evidence for critical periods in neuraldevelopment dates back to the classic experi-ments of Hubel and Wiesel4, who deprivedkittens of visual input to one eye and thenshowed that the animals’ binocular vision, as well as the responses of their visual-cortexcells, differed from those of kittens raisednormally. But identical manipulation ofvisual information had no effect in oldercats. Other familiar examples of critical peri-ods include language acquisition in humansand song learning by certain birds5. Criticalperiods in cognitive development are therationale for playing Mozart to babies.

Linkenhoker and Knudsen1 now showthat the auditory map in older owls doesindeed cope poorly with a large shift of thevisual world. However, by shifting the visualworld gradually, in several small steps, themap achieves a much greater degree of adap-tive change than was previously believedpossible. If, instead of being exposed to a single prismatic shift of 237, older owls werefirst exposed sequentially to shifts of 67, 117and 177, they acquired new auditory mapsthat could be reproduced by subsequentexposure to a single large prismatic shift.

These results by no means overturn theprinciple of a critical period. The auditorymaps in juvenile owls adjust to much largerchanges, and to about twice the extent, compared with adult owls. However, these

news and views

258 NATURE | VOL 419 | 19 SEPTEMBER 2002 | www.nature.com/nature

They say that you can’t teach an old dognew tricks. But on page 293 of this issue1

Linkenhoker and Knudsen provide evi-dence that you can at least teach an old owlnew tricks — if you train it in small steps. This result implies that the brains of olderanimals, including ourselves, may be capableof greater change in the form of adaptiveplasticity than previously expected.

Owls are nocturnal hunters which rely on auditory as well as visual information totrack their prey. Coordinated maps of audi-tory and visual space are formed in the optictectum, the midbrain region responsible fororienting the owl to its target2. The visualmap is essentially a direct representation ofinformation from the retina, but the audi-tory map results from a more complex com-putation, based in part on the microseconddifferences between the time of arrival ofsounds at each ear. For proper coordinationof its auditory and visual worlds, the youngowl must learn, for instance, that when it islooking directly at a mouse, the mouse’s ter-rified squeak reaches both ears at the sametime. But when the squeak reaches the rightear first by a certain number of microsec-onds, the owl should expect to see the mousea certain number of degrees to the right.

Experiments with juvenile birds haveshown that visual information has the upperhand in maintaining the coordination be-tween these two maps. In such birds, Knudsen and a colleague3 previously foundthat an optical shift of the visual world,achieved by placing prism glasses on theowls, results in a gradual shift of the auditory

map to keep sight and sound coordinated.This plasticity may seem somewhat mal-adaptive under the conditions of the experi-ment — that is, although it is the visual worldthat has changed the visual map stays thesame. However, it implies that outside thelaboratory, over the course of evolution,vision has proven to be the more accurateand reliable indicator of spatial location forguiding orienting behaviours.

Although adaptive adjustments in thebrain’s representation of the auditory world

Neurobiology

Plasticity and the older owlHemai Parthasarathy

Age reduces the brain’s ability to adapt to change. But a surprisingmeasure of neural adaptation does occur, at least in one experimentalsituation, if change is introduced bit by bit.

Figure 1 Adult and baby barn owls.Malleability of the representation ofthe auditory world in adult brains isgreater than expected.

AN

NE

KN

UD

SEN

& M

ELA

NIE

CIF

ER

RI

© 2002 Nature Publishing Group