Robert J. Geller (BS '73, PhD '77), Seismologist
Can earthquakes be predicted? From a public policy perspective, this would be a towering achievement in the field of seismology. With sufficient advance warning, lives and property could be saved. At various points in history, some experts have expressed optimism that earthquakes could be predicted, but the overwhelming consensus today, built over the course of the past fifty years of research, is that earthquakes cannot be predicted. Robert J. ("Bob") Geller has been at the center of these debates, and he is a leading voice in explaining why earthquakes cannot be reliably and accurately predicted.
The impossibility of earthquake prediction begs explanation. Is this impossibility a reflection of the inadequacy of our models, theories, and observational technologies—which would then leave open the possibility of some future breakthrough? Or, are the underlying geophysical systems that cause earthquakes fundamentally chaotic, which would suggest that the Earth itself does not "know" when an earthquake is about to start or how large it will become before it ends? In the latter scenario, no technological or intellectual advance would get us closer to predicting a phenomenon that is fundamentally unpredictable.
In the following conversation, Bob Geller discusses his education at Caltech, and the broad education he received in seismology and geophysics from undergraduate years through graduate school. He recalls some of the key topics of interest at the Caltech Seismological Laboratory ("Seismo Lab") in the 1970s and his thesis research in quantitative seismology. After a yearlong research fellowship at Caltech and six years on the faculty at Stanford, Geller was named Associate Professor of Geophysics at the University of Tokyo—the first tenured foreign professor in the history of the university. From his vantage point in Tokyo, Geller participated in some of the debates in Japan about earthquake prediction, and he reflects on how he assumed a role as a media commentator, public intellectual, and widely-read author in explaining why earthquakes cannot be predicted.
Beyond the specific topic of prediction, Geller has pursued a wide ranging research agenda, and his topics of interest range from studying earthquake source parameters to examining the causes of tsunamis. In the realm of public policy, Geller has been critical of the Japanese government's issuance of long-term probabilistic forecasts, which, he contends, are worse than being incorrect, because they can lead to the false assumption that certain areas at certain times are safe from earthquakes.
As Geller indicates, it is impossible to predict the future, and no one can dismiss with complete certainty the possibility that some future development could get us closer to earthquake prediction. But science hinges on the analysis of the best evidence available to us, and on that basis, Geller sees little reason to think such a breakthrough will happen. In 2017, Geller became professor emeritus, and his honors include becoming a fellow of the Japan Geoscience Union that same year.
[Ed. note: the bracketed numbers in the text correspond to the endnotes, listed at the bottom of the page.]
DAVID ZIERLER: This is David Zierler, Director of the Caltech Heritage Project. It's Thursday, April 14, 2022. I am delighted to be here with Professor Robert J. Geller. Bob, it's great to see you. Thank you for joining me today.
ROBERT GELLER: My pleasure.
ZIERLER: To start, would you tell me your title and affiliation?
GELLER: I'm Professor Emeritus at The University of Tokyo.
ZIERLER: Big question for your career, why did you decide to pursue an academic career in Japan?
GELLER: You probably know Hiroo Kanamori, who's now professor emeritus at the Caltech Seismo Lab. Hiroo joined the faculty of Caltech as a full professor in December of 1972, when I was a senior in geophysics. He was never officially my advisor, but he was a very big academic influence on me. Through my connection to him, I was head-hunted by people in Japan who were looking to hire a foreign faculty member. In 1984, I became associate professor at the Faculty of Science of the University of Tokyo. I was promoted to full professor in 1999, then I retired in 2017, which is mandatory at age 65.
ZIERLER: Did you know Japanese at all before you got to Japan?
GELLER: Yes, because I knew I was going to go there, so I did some studying up before I went.
ZIERLER: Were you a bachelor at the time? Were there family considerations in relocating to Japan?
GELLER: I was married at the time to an American woman, actually a Caltech graduate. She got her master's there. We were slowly drifting apart anyway. I was hoping she would join me in Japan, but she didn't. We were divorced several years after I moved to Japan, in 1987.
ZIERLER: Being in Japan for so long, what are some of the different cultural approaches to seismology that you see in Japan, even at a policy level?
GELLER: I don't think there are too many as far as the science goes. As far as public policy relating to earthquakes, maybe one big difference is that the capital of the United States is Washington, D.C., which had a magnitude 5.5 or so earthquake a few years ago, which caused a momentary flurry of interest, but is basically aseismic. But here in Japan, the capital, Tokyo, and the whole country, is earthquake prone. Earthquakes are a bigger agenda item.
A Japanese Perspective on Earthquake Prediction
ZIERLER: Why is it that some people in Japan believe that earthquakes can be predicted, and why do you argue with that?
GELLER: People here don't really believe that. In every country, in every field, obviously, scientists want to get government funding. In almost every country at almost every time, government funding is weighted towards things that promise practical applications. Even if there really are no practical applications, scientists will fake it. And even if there are some practical applications possible, but a long way down the road, scientists will tend to play that up. For example, in U.S. biomedical research, a lot of fundamental research is hyped as maybe having the potential to lead to a cure for cancer somewhere down the road. The same thing happens in many countries, including Japan, in many fields. In order to get funding for a program of observational science, some seismologists in Japan back in the 1960s overhyped the applicability to earthquake prediction. And they were actually very careful, they didn't lie. What they said was really amazing bureaucratic doublespeak.
They said, paraphrasing somewhat, "If you give us research funding for ten years, and our program is completely funded, there's a pretty good chance we can tell you in ten years whether or not earthquakes can be predicted." This gave them multiple outs. This was in the famous "blueprint" for earthquake prediction research published in 1962.  One out they left themselves was that no research program is ever wholly funded, so they could say ten years later, "Sorry, we can't answer the question because we were only partially funded." A second is, they didn't promise to say for sure ten years after the funding started whether or not earthquakes could be predicted, they just said they probably could. They obviously weren't optimistic that earthquakes could be predicted, but they had to play these bureaucratic games to get funded, just like everyone everywhere else.
Incidentally, there have also been efforts in the US to get funding for earthquake prediction. One example is a meeting in 1995 in Irvine, in January, I think, organized by the National Academy of Sciences, where a group led by Leon Knopoff from UCLA was trying to get US funding for earthquake prediction.  It was right after the 1994 Northridge earthquake. They planned the meeting before the Kobe earthquake in Japan. They didn't get much traction, but there is a group of people in the US who are continually trying to rev prediction efforts up again. It's not like all this is just taking place in Japan
This is a kind of puffery that goes on everywhere, although it probably shouldn't. The Nobel Prize-winning physicist, Richard P. Feynman, who I met a few times in passing when I was a student and postdoc at Caltech and he was a professor, was really big on telling scientists not to lie or exaggerate to get funding. On the other hand, he didn't turn down use of the funds and facilities that the physics department had obtained, perhaps partly by exaggerating the potential applications of their basic research. When I was a grad student at Caltech, I did a bit of work with one professor who's now deceased. I won't mention his name. But Seth Stein and I were working with him on a project aiming to measure the state of stress in the crust. The professor wrote a grant application promising all kinds of things. I think Seth and I were both a little skeptical. We said, "Hey, you can't say all this stuff."
ZIERLER: Why were you skeptical?
GELLER: Because he was over-promising things that were highly unlikely to be obtainable, particularly within the framework and resources of the requested funding and time period. He looked us right in the eye and said, "You guys don't understand. First, we have to get the money." I'm not making it up. There's a lot of that going on all over. That's how the "earthquake prediction" program in Japan started. You can criticize that, and in fact, I have. But at some level that goes on almost everywhere in almost every field; it's a question of research ethics that people don't really talk about. Research ethics is big on not faking data and stuff like that because that's done by the peons. But there are no rules against exaggerating things to get funding because that's done by the big guys. Anatole France said, "The law, in its majestic equality, forbids rich and poor alike to sleep under bridges."
ZIERLER: How did earthquake prediction efforts develop in the 1970s and later?
GELLER: What happened was, there was a kind of perfect storm in Japan, during the mid-1970s. Some seismologists were hyping up the supposed imminence of the "Tōkai earthquake" based on the idea that earthquakes occur in cycles. The Tōkai district runs along Japan's Pacific Coast, from the Izu Peninsula to Nagoya.
There was a seismologist who, at the time, was in the lowest-ranking faculty position at the University of Tokyo, which was then called research associate in English but is now called assistant professor, although it's not like an American assistant professor. He was going on TV and appearing in magazines, saying that a magnitude 8 earthquake "could happen tomorrow" in the Tōkai district. I was still at Caltech at the time, but his hyping of this in the Japanese media got sensational publicity and caused a quasi-panic about the supposed imminence of this quake. And there were more senior people in Japanese academia saying the same thing based on the supposed periodicity of earthquakes. Of course, Don Anderson at Caltech and Bruce Bolt at Berkeley, at that same period in the 1970s, hyped up the danger of a supposedly imminent quake ("the Big One") on the San Andreas Fault in Southern California. As we all know now, 40-some years later, nothing has happened there either. The zeitgeist of academic seismology all over the world then was making medium- or long-time scale predictions of supposedly imminent earthquakes based on periodicity. Since few if any of them have happened 40 years later, there's probably a scientific lesson in that for us. But let's leave that for later and get back to the mid-70s.
Based on the hype of the supposed imminence of the Tōkai earthquake, a few scientists got together with politicians in Japan and got the parliament of Japan to pass a law allocating huge amounts of money, to set up a system for the prime minister to be able to actually issue a warning to the public that a magnitude-8 earthquake would occur within the next 3 days in Tōkai; the issuance of this warning would result in evacuations, closures of highways and schools, etc. That's a step or two beyond anything the US ever did. When that system took effect, maybe it engendered a belief among the ordinary people in Japan that earthquakes were really predictable. It's probably a lot like when the Bush Administration in the early 2000s said that Saddam Hussein had weapons of mass destruction. They said it over, and over, and over again, and it sort of took root for a while in the popular consciousness, and the media were happy, for the most part, to go along. The people who opposed that, like the Dixie Chicks, were marginalized.
In Japan, in the late 1970s, there were several seismologists who were highly skeptical about the whole prediction thing. One of them was Hitoshi Takeuchi at The University of Tokyo. He was a well known seismologist maybe one generation before Hiroo. When Takeuchi had just become a full professor, Hiroo was hired in his lab as a research associate in, I think, 1962, and he had that position until 1966. He was on leave from that position when he was a post-doc at Caltech from 1964 to 1966. Anyway, when Takeuchi became a full professor, he was Carl Sagan before Carl Sagan. He more or less stopped doing research and started being a TV personality, writing books, doing that sort of thing. But he was publicly very skeptical about earthquake prediction. His views were reported in the newspapers, but the next day, they went back to reporting on the official government line, that earthquakes could be predicted.
When I came to Japan in 1984, every year on September 1, which is the anniversary of the 1923 Tokyo earthquake, they had a little ritual on TV, where the six members of the committee that was tasked with being convened if supposed precursory anomalies were observed, were whisked, as a drill to simulate the real thing, by the police with sirens wailing and lights flashing, to the headquarters of the meteorological agency. Then, they had a mock meeting to review mock anomalies. They duly said, "Hey, these are anomalies," and a mock notification was made to the prime minister, who issued a mock proclamation that the Tōkai earthquake was going to happen within three days. And this was all on TV, on the semi-government channel NHK. And they did that every year. This mock drill that assumed the Tōkai earthquake could be predicted went on every year until 1995, when the Kobe earthquake hit. I'm not sure exactly when, but shortly after that, the mock drill was taken off TV, even though they continued to carry it out for a while. I think they were sort of embarrassed by the whole thing.
Finally, in 2018, they sort of fessed up in very obscure language that the imminent prediction of a large earthquake wasn't really possible. The law authorizing the prediction to be made still exists on paper, but they've toned down what they're doing. Now, what they do instead, this is all a charade, and they know it's a charade, and everyone knows it's a charade, but in order to keep their funding, they've toned their activities down to the idea rather than making a definite prediction that the Tōkai earthquake will occur within three days, they just will make a statement that there's a somewhat greater than normal chance that a great magnitude-9 earthquake in the Nankai trough region, running from Shizuoka, where the Tōkai earthquake was supposedly going to occur, all the way to maybe the western tip of Shikoku, the smallest of the four main Japanese islands, will occur in the next week or so. There have been large earthquakes at the Nankai trough in the past, some along the entire length, and some along only parts of it. There are obviously going to be more in the very near geological future, but that means hundreds or even thousands of years. We'll obviously have a number of great earthquakes there on that time scale in the future, but maybe not any time soon, on a human time scale.
Under the present system, if there's a magnitude-6.7 or larger earthquake in the Nankai Trough area, the six-person meteorological agency advisory committee will issue a statement saying, "It's maybe somewhat more likely than normal that a magnitude-9 earthquake will happen tomorrow." Let's make up some numbers. Say, there's going to be a magnitude-9 earthquake like that every 200 years. Not necessarily along the whole segment, but perhaps along part of it. That's not a periodicity, just a rough average occurrence interval over geological time. That's 70,000-some days.
So let's say chances at random of a magnitude-9 Nankai earthquake tomorrow are 1 in 70,000. If the advisory committee tells you, "Because of the recent seismic activity, that chance has gone up a little bit. Now, it's ten times more likely than normal, 1 in 7,000 tomorrow rather than 1 in 70,000," what are you going to do? That's too low a gain in probability to do anything other than what you should've been doing anyway, like making sure you had a week of bottled water in your house and so on. Making sure that your bookshelves and cabinets were bolted to the wall or that you have these adjustable gadgets (see photo) to keep them stuck firmly to the ceiling so they won't fall over in an earthquake. Things like that, you should've done anyway. This is all just a charade for them to keep their funding and keep the advisory committee system from being abolished.
By now, the public and the media don't really believe anymore that earthquakes can be predicted reliably, within at most three days, before they occur. I suppose, what I've been saying to the public for the last thirty years or so may have had some effect.
A Philosophical Problem in Seismology
ZIERLER: I'll ask a question that's going to brush up even on philosophy. When you say that earthquakes are not predictable, to what extent is that a reflection on human limitations, or does the Earth itself not "know" when an earthquake is going to occur?
GELLER: That's a good question. The answer to that question is in two parts. The first is, what can we do right now, at this instant in time, based on present scientific knowledge? At the present time, we have no reliable and accurate way of predicting earthquakes, where we define prediction as an imminent warning of a large earthquake with enough specificity in magnitude, time, and place to justify issuing an alarm. I think all reputable seismologists or other scientists would agree to that statement. There are a few fringe scientists making all kinds of bizarre electromagnetic observations or whatever who claim they can make predictions now, but they're not taken seriously within the scientific community, and they have no real data to back up what they're doing. We can't do it now. The second of your questions is asking about the science. My personal feeling is that earthquakes are too complex and nonlinear a phenomenon for them ever to be predictable with enough accuracy and reliability for issuing alarms. On the other hand, that's speculation about how science will evolve. We can never rule out the possibility that someone will find something new in the future. But until that happens, we can say, based on the present scientific knowledge, it's not possible.
I have one more point I always make in that context, which is that if you go back 120 or 140 years to the dawning of seismology, interest in prediction has waxed and waned. But at some level, there's always been somebody speculating about earthquake prediction. None of those efforts over the past 140 years has panned out. On the other hand, in terms of what we know now scientifically about earthquakes, and in terms of our instrumentation for observing earthquakes, that's gone right off the charts over the last 140 years. We've made tremendous progress in basic science and instrumentation that didn't lead to any progress toward earthquake prediction. Make of that what you will. There are any number of other fields where you have very complex, nonlinear systems that make it either impossible to make predictions or put some severe limitations on the accuracy with which they can be made. For example, turbulence in fluid dynamics. Once turbulence starts to happen, all bets are off about how the system will evolve. That seems to be true of earthquakes. As far as whether 100 years from now, people will be able to predict earthquakes or not, my guess is no, but that's only a guess. I probably won't be around in 100 years to have to take responsibility if I was wrong. But I'll just put that on the record.
ZIERLER: In your career, have you been more on the theory or the observation side?
GELLER: Basically, my specialty is computational seismology, data analysis, and theoretical seismology. What I really do for a living, or did for a living, and am still working on for fun, is numerical modeling of seismic wave propagation in complicated, heterogeneous media, then analyzing data to infer the 3D structure of the Earth's interior. Not only P-velocities and S-velocities, but also seismic anisotropy. I'm still doing research and writing papers in that area, despite having retired. This earthquake prediction thing is just sort of a sideline that I pursue as a citizen. I got really tired of all this bullshit being propagandized to the public and I thought, "Someone has to say something to call bullshit," as it were. Back in the day when I was getting started, people couldn't use language like that in public. But thanks to President Trump, it's become acceptable now.
ZIERLER: Let's go all the way back to the beginning. Tell me about your initial interest in going to school at Caltech, coming all the way from New York. How did that happen for you?
GELLER: There was a guy from Caltech, I think named Jones, who was the director of admissions for Caltech. He would hit the dog-and-pony circuit throughout the country, going from one high school to another, making presentations about Caltech. I heard his presentation when I was a junior at my high school in New York. Of course, I'd heard of Caltech before that. I think any high school student interested in science has heard of places like MIT, Caltech, and so on. But hearing his talk got me a little interested, so I applied there. I liked the fact that it was small. There were only about 200 undergrads in each class when I went there. And having grown up in New York my whole life, I thought it would be fun to live in another part of the country.
ZIERLER: Were you interested in seismology and geophysics even before you got to Caltech?
GELLER: Not except very casually. I'd read a little about it. I was mostly interested in physics and math. I started at Caltech when I was 16, having skipped a couple of grades in elementary school. Then, I started majoring in physics. Actually, I was a little immature, so I decided to take a year off after my sophomore year, and I worked for an oil exploration company, just doing sort of grunt work, some of it related to seismological modeling of the Earth in connection with prospecting for oil, what they call exploration geophysics. When I came back after working for that company, there was a professor at Caltech, not in the Seismo Lab, but in the Division of Geological and Planetary Sciences, named C. Hewitt Dix. I looked him up, and he was very kind, and we had a nice chat. He gave me a lot of good advice. He was a math PhD at Caltech, and he got his PhD during the depression, when there weren't many jobs. Having gotten a PhD in math, he went to work for Gulf Oil in Pittsburgh in their research lab. He worked on seismology, and he made a lot of contributions to exploration seismology (i.e., seismic prospecting for oil and gas), wrote books and papers in that field that are still being cited today. He was a good example for me of someone who moved from basic research into a more applied field.
Having worked doing grunt work for an oil exploration company for the year I took off, I was interested in majoring in geophysics. Dix gave me some good advice. Actually, he'd been very friendly with Hiroo when Hiroo was a post-doc at Caltech in 1964-66. He gave me a present of a book in English about plate tectonics and continental drift called Debate About the Earth  by three guys. One was Takeuchi, the guy I mentioned who went Carl Sagan. A second was Seiya Uyeda, who I actually later met when I was a grad student at Caltech and post-doc. He was a visiting professor then. I'm sad to say he's gone over to the dark side now and is doing all kinds of highly questionable earthquake prediction stuff. Then, the third author of the book was Kanamori. The Japanese version of the book was only by the first two, then Hiroo was brought in as the third author when they did the English version. Dix made me a present of that book. That was in 1971. Maybe he knew that they were trying to get Kanamori to Caltech then, but he didn't mention that to me at the time. I became a geophysics major when I came back and started to work at the Seismo Lab doing programming.
ZIERLER: What kind of programming?
GELLER: I was working on assembly language, that sort of thing. David Harkrider, Charles Archambeau, and Don Helmberger were doing a lot of work for the Air Force and Defense Advanced Research Projects Agency on modeling seismic waves generated by underground nuclear tests and things like that. I would run programs for them or fix them for Harkrider, and they bought a minicomputer for the Seismo Lab, a Nova made by a company called Data General. I did programming for that, and I would sometimes take Harkrider's programs, which were on decks of punch cards then, and go from the Seismo Lab to the computer center at the main campus, run them on the IBM mainframe, stuff like that. That was kind of fun. That also got me to look at research as a career. The other thing is, my birthday is February 9. On my 19th birthday in 1971, the San Fernando earthquake happened. That was when I was on leave from Caltech. That was the first big earthquake I'd ever felt anywhere.
ZIERLER: Where were you when that happened?
GELLER: I was sharing a house with another guy in Pasadena. For a long time I'd forgotten the exact address and I thought it was Altadena, but I recently had occasion to look up a list of all the places I'd ever lived, and I discovered it was in northern Pasadena, not Altadena. The house really shook around, and I was a bit surprised, having never felt an earthquake before. An earthquake at the base of the San Gabriel Mountains was not what seismologists or geologists "expected" then. Even then, they were talking about "the big one" on the San Andreas Fault as the "expected" next earthquake. It's really a kind of amazing thing how the "expected" earthquakes almost never seem to happen, while almost all the big ones are "unexpected."
These days, the air is a little better in Pasadena, so you can see the San Gabriel Mountains every day. Back then, you could only see them sometimes. But the smog wasn't so bad in the winter, so you could see them then. When I first got to Caltech in September of '68, you couldn't even see them most days. But the San Gabriels, even to someone who's not super into geology, it was obvious that they were a recent geological feature. Something had to have caused them to be there. Not the tooth fairy, the mountain fairy, or whatever, but motion on faults. But somehow, most people were still "surprised" by an earthquake at that particular time in that particular place. That got me a little interested in earthquakes, too. But my main interest was in numerical modeling, that sort of thing.
Caltech the Whole Way Through
ZIERLER: Did you ever give thought to going somewhere else for graduate school? Or you knew you wanted to stay at Caltech?
GELLER: I thought about going to some other places. Tom Jordan was a grad student at Caltech when I was a junior. Being at the old Seismo Lab—it's probably different now on campus—everyone there got to know everyone else. Tom, who became assistant professor at Princeton when I was a senior, tried to convince me I should go to Princeton as a grad student, but that wasn't very persuasive coming from someone who, himself, had stayed at Caltech as a grad student after being an undergrad there. I also looked at some other schools, like UC San Diego or the University of Washington. But having worked at the Seismo Lab, I decided I should stay there. It was actually, in retrospect, a sort of stupid idea. It would've been much better, in principle, to go someplace else and be exposed to new stuff. I don't think I was very smart. But serendipitously, Hiroo Kanamori came to Caltech as a full professor in December 1972 during my senior year. Being able to work with him as a grad student was sort of like going to another place, so it worked out OK.
ZIERLER: Did you know who your thesis advisor was going to be?
GELLER: It was Dave Harkrider because I'd been working with him all along. He was officially my undergrad advisor and graduate advisor.
At the time, in China, they had the so-called Gang of Four under Mao Zedong, consisting of his wife and three other people. People would call three grad students, Emile Okal, Seth, and me, the Gang of Three. I didn't really think this was anything unusual at the time, but in retrospect, it sort of was. The three of us did a lot of research and wrote a lot of papers together, without faculty as co-authors. We did get a lot of advice informally from the faculty, especially Hiroo. Not only Hiroo, but some of the younger Japanese associates of his came as post-docs, and it was fun talking to them, too. But the three of us sort of worked mostly with each other a lot. It wasn't really your classical graduate student experience. Each of us got involved in writing proposals—grant applications—that went out under the faculty's names. Most of them got funded. From the faculty point of view, here were these students who were sort of taking care of themselves and even bringing in some money, so they pretty much left us alone.
ZIERLER: What were some of the big debates in the Seismo Lab when you started graduate school there? What were people talking about?
GELLER: One of them was the so-called baseline shift. Tom Jordan, in his doctoral thesis, looked at the Earth's modes of free oscillations, which are like long-period surface waves. From the measured periods of the modes of free oscillations he and Don Anderson derived an earth model, that differed significantly from previous earth models. There was some speculation that this discrepancy was because the free-oscillation data sampled the Earth uniformly, whereas the previous models reflected primarily the composition of the Earth under continents. However, what should've been known at the outset but was only worked out a bit later was that if you were looking at free oscillation modes with periods of 1,000 or more seconds, and you're comparing the earth model inferred from those to the earth model inferred from body waves, like the famous Jeffreys-Bullen model, based on body waves with a period of one or two seconds–there's a phenomenon called physical dispersion due to anelastic attenuation. This is very well-known in physics. It's why red light travels slower than blue light in the atmosphere, why a prism divides things up. If you want to compare an earth model derived from long-period observations to one derived from short-period observations, you have to correct for physical dispersion due to anelastic attenuation, which was initially not corrected for. There was a paper by Hsi-Ping Liu, who was a post-doc then, with Anderson and Kanamori, published in 1976 , where they made that correction. And the baseline shift went away. That was one issue.
Another issue, which is still sort of ongoing, didn't involve Caltech but a paper by Forsyth and Uyeda , who were then at MIT, about what the driving forces of plate tectonics were. When you get a paper to review, it's supposed to be a secret. The situation is often complicated, as in this case, because the authors had already presented their ideas in seminars and at scientific meetings, so their theory was public, even though their specific manuscript was not. Of course everyone respected their priority and no one at the Seismo Lab tried to rip off their work. But on the other hand, many people at the Seismo Lab were really interested in the topic. So these authors' theories were a lunchtime discussion topic for two or three weeks. The fruits of these discussions, in condensed form, were fed back by the reviewer to the authors, and I think this helped them to sharpen up their paper before it was published. It has about 1300 citations now, by the way. That kind of discussion was very educational for a young scientist. Lots of hot scientific issues were always being discussed like that while I was at the Seismo Lab.
Earthquake prediction wasn't really a major issue while I was at the Seismo Lab, although it was sort of a peripheral issue. There were some guys in the late 1960s in the Soviet Union, as it was at the time, in what's now the Republic of Tajikistan, in a place called Garm, and they claimed to have seen humongous changes in seismic velocities before earthquakes. They said that the in-situ seismic velocity went way down, then back up, then you had the earthquake.
This was just crazy stuff. But somehow, it took on a life of its own. In 1973, there was a paper by Jim Whitcomb, Jan Garmany, and Don Anderson . Whitcomb was a post-doc at Caltech at the time, and Garmany was an undergrad student, my classmate, who worked at the next desk to me in the old Seismo Lab. Anderson was then the Director of the Seismo Lab. More or less simultaneously, another group at Columbia University's Lamont-Doherty Geological Observatory  published similar papers also claiming to have seen these humongous velocity anomalies before earthquakes. None of these were really changes in actual in situ seismic velocities; they were changes in apparent velocities, which is another animal altogether. Clarence Allen and Don Helmberger  measured real velocity changes from quarry explosions and found there wasn't much of anything. From my vantage point as a young grad student, that was a pretty effective debunking effort. But in public, the supposed velocity anomalies were really hot stuff in terms of publicity.
ZIERLER: When you say that people weren't really working on earthquake prediction at that point, is that because it wasn't taken seriously or because we simply didn't have the tools to study it in a rigorous way?
GELLER: Mostly the former. I once asked one faculty member, now deceased, with whom I was pretty friendly–we would have a drink together from time to time—what he thought about earthquake prediction. He told me, "It's a nice hobby, but I like women and whiskey better." He really said that. But on the other hand, he and the other faculty didn't rain on the parade in public.
ZIERLER: Because the public demands prediction for earthquakes?
GELLER: No, it's like doctors. Maybe everyone in the county medical society knows that Smith is a real butcher, but they'll never say so in public. There's a professional code of omertá. But if you read between the lines of the paper by Allen and Helmberger about the absence of velocity anomalies in the quarry explosion data, it didn't take a super genius to figure out what they were saying. But they didn't actually say it, except maybe in a very circumspect, indirect way. Anyway, this was a worldwide phenomenon. In my 1997 paper reviewing the history of prediction research , I quoted a number of the optimistic statements made by people in the USGS and elsewhere in the 1970s. I was at the next desk to Jan Garmany, so I'd seen the data. I couldn't believe that this was the stuff that was going to get them the Nobel Prize or whatever. Having been brainwashed by all the stuff Feynman said about how scientists should always tell the truth, I probably would've said that to a reporter, if anyone had ever asked me, but I was an undergrad, and then a first year grad student, so no one from the media ever asked me.
ZIERLER: How did you work on developing what would become your thesis topic?
GELLER: I just wrote a bunch of papers, then I got a stapler.
ZIERLER: What were some of the commonalities in those papers?
GELLER: They got published. [Laugh] I said this in a light-hearted way, but Seth and Emile and I were just trying to do interesting research and write and publish good papers on topics that interested us, and we really didn't restrict ourselves to working linearly towards a predetermined thesis topic.
ZIERLER: [Laugh] I meant substantively.
GELLER: I'm being honest. There was one part about earthquake source parameters and another about the earth's free oscillations. They're both quantitative applications in seismology. But they were just sort of interesting. Probably I could've made a thesis out of either half of them, but what the hell.
Modernizing the Magnitude Scale
ZIERLER: What were some of the key conclusions you arrived at in graduate school?
GELLER: One of them was actually building on a paper by Kanamori and Anderson published in 1975 about scaling relations in seismology . That means plotting various parameters on a log-log scale and looking at the slope of the relations. One of the things that got straightened out, largely due to Kanamori, although I perhaps made some small contribution, was the difference between earthquake magnitude (as defined by Richter and Gutenberg) and the real size of an earthquake. How much have you gone back and read about Richter and Gutenberg, and how they developed the magnitude scale?
ZIERLER: Quite a bit, but please, tell me.
GELLER: The impetus for that was inquiries from the media. After the 1933 Long Beach quake, they'd get phone calls every time there was an aftershock. "How big was that?" They had to come up with something, and Richter came up with the magnitude scale based on magnitudes in astronomy. Then, later on, he and Gutenberg tried to relate earthquake magnitudes to earthquake energy, how much energy an earthquake released. But the seismometers at the time had limited bandwidth, which only let them go up to 20- or 30-second periods. Benioff, also at the Seismo Lab, made a longer-period seismograph in about 1950 that proved very useful. Anyway, the 20-second surface-wave-based magnitudes of Richter saturated at about 8.5. The reason for that was destructive interference between waves, that the earthquake goes on for five or ten minutes if you have a really big one. If you have an earthquake going for that long, the 20-second-period waves will destructively interfere. That's why the magnitude, as measured by a 20-second instrument, saturates. Seismic moment, which is a theoretical parameter in earthquake source theory, defined as the product of the average rigidity—that is, the shear modulus of the rocks surrounding the fault zone—the fault area, and the average displacement, is a better measure of the earthquake size. My work on the effect of the earthquake fault width  contributed to understanding that saturation process of Richter's magnitude formula a little bit. Then, finally, right around the time I was leaving Caltech, Kanamori proposed taking the seismic moment and using an empirical formula to backwardly turn it into a magnitude .
ZIERLER: What does that mean, the seismic moment?
GELLER: I just defined it in words, but as a formula it's M0 = μSD, where M0 is the seismic moment, μ is the rigidity of the rocks surrounding the fault zone (in units of dyn/cm2), S is the fault area (in units of cm2), multiplied by the average slip on the fault in the earthquake, D (in units of cm). M0 has units of dyn cm, which may appear to be the same as the units of energy. But seismic moment is not an energy. It's actually about 10,000 times the energy, very very roughly speaking. There were actually some famous guys who, in the mid-70s, screwed up and forgot that. Just like in any other field, it's easy to make mistakes, but they get caught in a hurry.
The US installed a global network, the World-Wide Standardized Seismograph Network (WWSSN), in the 1960s. The purpose of this was to make data publicly available for lots of things. But from the US government point of view, they wanted to understand seismic waves generated by underground nuclear tests and how to discriminate them from earthquakes..
When you had a really big earthquake, like the 1964 Alaska earthquake, the WWSSN instruments saturated. They went off-scale. But if you waited a couple hours, the surface waves that had gone all the way around the Earth once or twice were back on scale because their amplitudes were diminished by attenuation as they traveled. When Kanamori was still in Japan, he showed how you could measure the seismic moment from these long-period surface waves on seismograms. He showed examples of measurements for two large earthquakes in his two famous papers in 1970 .
By the end of the 70s, he'd developed a new formula for the "moment magnitude" of an earthquake, which is denoted by the variable MW. The moment magnitude is obtained by taking the measured seismic moment, and converting it to a "pseudo magnitude" using an empirical formula proposed by Kanamori . Using MW to quantify the size of earthquakes eliminates the saturation problem. While I was at Caltech as a post-doc, my only paper written together with Hiroo was one where we went back through Richter's paper notes, and we tried to figure out how he had actually measured magnitudes . Richter didn't always describe the algorithm he used to combine body-wave and surface-wave measurements into a single parameter.
ZIERLER: Who was on your thesis committee?
GELLER: Harkrider, Kanamori, Helmberger. I think maybe Jerry Wasserburg. Maybe Don Anderson.
ZIERLER: What opportunities were available to you after you defended?
GELLER: I was a post-doc at the Seismo Lab for a year, then I got head-hunted by Stanford, where I went as an assistant professor for six years. Then, in 1982 or so, I was head-hunted by the University of Tokyo. Through my association with Hiroo, I was sort of honored to be head-hunted by his alma mater.
ZIERLER: What was the research you were doing prior to getting recruited by Tokyo?
GELLER: Basically, the same stuff I did after I went there. Working on the Earth's free oscillations and analyses of seismic data of various sorts. I was always opportunistic. When people around me found interesting things, I did some work with them on that. For example, it was never a long-term project of mine, but I wrote one paper in 1980 called Four Similar Earthquakes on the San Andreas Fault about events that occurred at different times, but had nearly identical waveforms . That short report by myself and Chuck Mueller, who was then a student at Stanford working with the USGS, is one of my most-cited papers. Some people now call those earthquakes "repeating earthquakes," or "repeaters," but, strictly speaking, they're not really repeating. They look the same at long periods, but at higher frequencies, they're different. They're not exactly the same earthquake repeating, but they're earthquakes in very nearby places. That was my only paper on that subject, but it's still continuing to be cited now. It's up over 200. Mainly, I was working on the Earth's free oscillations and the structure of the Earth, methods for computational seismology and so on.
ZIERLER: Were you collaborating with Hiroo, or you were just generally in touch with him at this point?
GELLER: The latter.
ZIERLER: Did you mostly work on your own? Who were some of your key collaborators at that point?
GELLER: All my students at Stanford at the time, I wrote papers with almost all of them. Also, Seth Stein was a post-doc at Stanford for a year or two before going to Northwestern.
The Opportunity in Tokyo
ZIERLER: What is your sense of why the University of Tokyo was specifically looking for foreign-born scholars?
GELLER: They weren't at all as an institution, but there were a few guys in geophysics who wanted to hire a foreign professor. It had become legally possible in 1982. At that time, the University of Tokyo was still a national university. Since 2003, it's become a corporation under the government, kind of the way the University of California is under the regents corporation rather than being directly under the state. But at that time, it was a national university, and all the faculty were civil servants. In almost every country in the world, it's hard to hire foreign citizens as civil servants. A special law allowing universities to do that passed in 1982. I mentioned Takeuchi to you. When he retired in 1981, some of the people around him wanted to shake things up a bit. One way was to bring in a foreign faculty member. These two or three guys proposed my hiring, and probably, the university as a whole is a bureaucracy like Caltech or anywhere, where you need someone championing something to get it done. Otherwise, it just goes on in a straight line. I was hired as a tenured associate professor and started there in August 1984.
ZIERLER: When did you start taking language classes? How much preparation did you have?
GELLER: I was still a faculty member at Stanford, and they were very nice. The guy teaching the class let me sit in on the intensive summer ten-week course in the summer of 1983. Every morning for ten weeks from 8 til 12, I'd go to the lectures, then after class I'd listen to the cassette tapes over and over. I acted like a real student and took the tests too. That's the only formal education in the language I ever had. It obviously wasn't enough to become proficient. But on the other hand, it gave me a good platform of basics to build on. It was really important.
ZIERLER: Was this a tenured offer?
ZIERLER: Is the tenure system in Japan similar to that in the United States?
GELLER: First of all, let me answer about how it was at the time, and then what it's like now. It's changed quite a bit. When I was hired, everyone was a civil servant. Everyone with Japanese citizenship was hired with tenure, but not by virtue of academic accomplishments, just by virtue of the fact that all civil servants had permanent jobs. You had to really screw up to get fired, maybe murder somebody, get arrested for drunk driving, something like that. Otherwise, you were in. Even the research associates. There were many people who were hired as research associates and spent their entire career at that level without ever being promoted, and they retired as that. Usually, unless someone was completely indolent, for example, a research associate from the University of Tokyo could become associate professor at a lower-ranking university. But there was no official up or out system. There still are probably a few people left over from the old system. What happened in a lot of departments, although not mine, was that there was a kind of informal agreement that even though you officially were permanent, you would be there only five or ten years as a research associate, then would go somewhere else. That worked out pretty well in physics, where they hired reasonable people, and every university in Japan needs physicists. Everyone was expected to leave.
Now, since becoming a corporation under the government, there are many officially untenured or tenure-track positions. Research associates are usually hired for X years nowadays, three years with a possibility of renewal for three years or whatever. Although, in my former department, when I retired, they were still hired permanently, but with an informal understanding that they would leave within, at most, ten years. Now, there are more and more associate professors with tenure-track appointments. It varies among universities. They haven't yet come up with a completely standardized national system.
At the time I was hired all Japanese faculty at public universities at any level were tenured, whereas foreign faculty at any level could be either tenured or appointed for a fixed-term (renewable by the employer, but with no guarantee). When I was hired in 1984 as a tenured associate professor, I was actually the first tenured foreign faculty member in the history of the University of Tokyo. In 1990 I wrote a short article about the problems in the system for employing foreign faculty . Things in Japan have changed a lot since then, and there now isn't much difference between the system for hiring Japanese and foreign faculty.
ZIERLER: How big was the program when you joined?
GELLER: When I was hired, I was hired by the Geophysics Department at the University of Tokyo. At that time, in my department, there were 20 faculty members, more or less. They were in five sections, three for solid Earth or planetary geophysics and two for geophysical fluid dynamics, one for physical oceanography, and one for meteorology. Each section had a professor, associate professor, and two research associates. In addition to that, there were three other Earth-science departments in the Faculty of Science: Geology, Mineralogy, and Physical Geography. They were completely separate departments and in another building. In addition to that, there's something called the Earthquake Research Institute (ERI), is another independent institute within the university. ERI isn't all geophysics—it also includes some earthquake engineers. I can't remember how many professors there were then at ERI, but recently, there were, like, 28 professors, 28 associate professors, many technicians and research associates. It's humongous.
ZIERLER: Was the faculty focused on the same kinds of things that faculty in the United States were working on?
GELLER: Yeah, they were trying to get grants. (That was a joke.) For the undergraduate education, it was mostly carried out by the people at the Faculty of Science. But there's not only an Earthquake Research Institute, there was also an Ocean Research Institute. The Institute for Solid-State Physics had a group working on high-pressure experiments. All of these different groups, some of which I haven't mentioned, collectively carried out graduate education in geophysics. When I was hired, each of the four undergraduate departments had its own graduate program with core faculty members from the respective undergrad departments in the Faculty of Science, but also with participation by faculty from the Earthquake Research Institute, and so on. In the year 2000, the four graduate departments were merged into a single graduate department. But there still were two undergraduate departments, one for geophysics and planetary physics, and one for geology, in a broad sense. These are still somewhat separate now. It's very siloed.
The emphasis of research is probably different than in the US, especially in solid Earth geophysics. Because of the funding nominally allocated to earthquake prediction until 1995, and after 1995, to basically the same group of former earthquake prediction guys, but with a slightly different name, there's a huge number of people in those fields. They're mostly doing competent routine observations that are represented as having the potential to contribute to earthquake prediction, or to earthquake forecasting, or whatever. There's a highly variable rate of production from one lab to another in terms of publications, citations, or whatever. There's a lot of mostly competent but not very interesting stuff as well as some interesting stuff.
ZIERLER: Did going to the University of Tokyo change the research you were doing?
GELLER: Not so much, but maybe a little. In terms of my work on long-period seismology and numerical modeling, there was a natural progression, but you can trace back what I'm doing now to what I was doing at Caltech as a grad student. Although it's evolved quite a bit. On the other hand, when I was watching all this earthquake prediction stuff, I got very worried. People were saying all kinds of stuff I thought was unreasonable. Also, the way the funding was allocated to "earthquake prediction" really distorted the science. That's still a problem now. I was successful in convincing the public, the government, and the media that you couldn't make reliable earthquake predictions. But basically, bureaucracies have a remarkable ability to survive by changing names, programs, and so on, without really changing the people running them. That's happened now.
The media in Japan are still continuing to report that the so-called Nankai Trough great earthquake is imminent, that it has an 80% chance of happening in 30 years. That's just bullshit. It's not a scientifically testable statement. If you have some scientific theory that lets you predict the probability of earthquakes in every seismic zone based on alleged periodicity, and you wait 10, 20, 30 years, or whatever, you can then test the success rate of those forecasts statistically. But if you have only one statement about one seismic zone, it's not really testable. Either an earthquake happens or it doesn't in that area in 30 years. But that doesn't prove anything. It could've just been a coincidence. Every year, they make the same kind of announcement. They started making them in 2003. It's already 20 years with no earthquake. I got involved in debunking work in that field just because I couldn't stand watching it and saying nothing .
ZIERLER: Did you experience any backlash as a result of this?
GELLER: Surprisingly little. Maybe it's because I'm a foreigner, or maybe it's just because I have a thick skin, and I'm insensitive. But nothing very bad happened. I started doing this when I was an associate professor, and I got promoted to full professor in 1999 anyway.
ZIERLER: But you were debunking things on a political level, not a scientific level.
GELLER: Well, both. There's this guy, Panayiotis Varotsos, who was a professor of physics at the University of Athens in Greece, who has been claiming since 1980 that he can predict earthquakes. In the early 1990s he managed to get tremendous publicity for his claims of making successful predictions in Greece. He tried to get funding from the UN's International Decade for Natural Disaster Reduction (IDNDR), and a conference was held in London to evaluate his claims . In my paper for the conference  I went through all his predictions, and I found most of them were "successful" only by retrospective moving of the goalposts (i.e., fudging or fuzzing up the criteria for "successful predictions"). Also, I attacked his publicity tactics. He even got his Japanese friends to publish a comic book in Japanese saying that Varotsos could predict earthquakes. I thought that he shouldn't be allowed to make these claims without some pushback, so I decided to get involved. I got sort of sucked into spending much more time on this debunking exercise than I originally envisioned.
ZIERLER: How much were you able to convince the people who needed to be convincing that in fact earthquakes are not predictable? How much success did you have?
GELLER: In terms of Varotsos, he was trying to make an end run around the normal funding system by getting a UN program (IDNDR) to endorse his work and get direct funding outside the normal review system. And in the end they didn't fund him. Not only I, but also all the other people criticizing him, managed to do a good job. He's still around, and he's still publishing stuff, but he doesn't really get any traction. Basically, in terms of earthquake prediction, the scientific community as a whole accepts that it's not possible at this time to make reliable and accurate predictions of imminent (say three days or less) earthquakes.
The second part of the controversy is the long-term earthquake forecasts based on the idea that earthquakes occur in cycles. That's something where I'm not the primary critic of those methods. One of my good abilities is, I'm probably a much better writer than your average scientist. I frequently get involved in projects talking to people, then my involvement is not only in the work, but in helping getting it written, published, and so on. David Jackson and Yan Kagan at UCLA, I sent you a paper last night by them and myself. In 1997, there was a paper in Science called Earthquakes Cannot be Predicted by myself, Jackson, Kagan, and my friend Francesco Mulargia from the University of Bologna . The four of us were also involved in various ways in the debunking of Varotsos's work.
Jackson, Kagan, Mulargia, and myself got to be friendly with each other. I'd known them before, through the Varotsos debunking. I became critical of Varotsos's prediction claims, but at the outset I knew almost nothing about them. I was one of the editors of Geophysical Research Letters, and decided to publish a special issue  where the critics could criticize and Varotsos and his supporters could reply, and then the whole scientific community could make up their own minds. I was the editor putting that together. I wasn't fully aware of all the problems when I got started. My initial involvement was through Seiya Uyeda of the Earthquake Research Institute of the University of Tokyo, who I'd met at met at Caltech, who was a big fan of Varotsos. And as editor of the journal Tectonophysics, Uyeda published many of Varotsos's claims to be able to predict earthquakes. Frankly, I hadn't really paid much attention to them. But in the early 1990s, they were popping up in the Japanese news media. His slogan was, "60 or 70% success rate of predictions." And that was all just by goalpost-moving of the definition of success.
Varotsos was invited by Uyeda to give a seminar at the University of Tokyo, which I heard, and it was a bit hard to understand and seemed a little fishy. I got involved in exploring it. I gradually understood what all the pitfalls were. Another thing, I was watching the cold fusion debacle  from the sidelines, but it was really interesting to see how those bizarre claims got a lot of traction in the popular media. Just out of a sense of curiosity about what was going on around me, I started to get involved in evaluating, and then debunking, Varotsos's claims.
As I mentioned, Knopoff organized the 1995 Irvine meeting, and there was a pro-prediction article in Science by Paul Silver and Hiroshi Wakita in 1996 . I think the Science people were a little bit ashamed of having published it. In an attempt to be fair, they invited me and the other guys to write the earthquake prediction debunking paper. That's the origin of our 1997 Science paper. Then, there was a sort of anti-Irvine and anti-VAN meeting in 1996 in London, which led to a special section of an issue of Geophysical Journal International  in 1997. There are articles by Kagan, myself, Mulargia, and Philip Stark from Berkeley, as well as many others.
I think it's fair to say that you still have people claiming that they can observe anomalies before earthquakes that maybe could be used for prediction, and those guys are continuing to publish, but they're basically a fringe group not getting any traction. It's a problem, but not such a serious problem. The more serious problem is the issue of whether earthquakes are occurring in cycles or not. The term "the earthquake cycle" has become so entrenched in the way seismologists or geologists talk, despite the fact that objective testing shows earthquakes aren't cyclical. Kagan, Jackson, and I  have pointed out the problem of the continued widespread use of "earthquake cycle" (as well as other similar terms). Then, Seth Stein and Mian Liu at the University of Missouri, along with myself, published a paper in 2012  about the problems of earthquake hazard maps. And in a related paper in 2017, Mulargia, Stark, and I  pointed out other problems with using the cycle models to forecast probabilities, set insurance rates, and things like that. Our papers have been well-cited, but they haven't really had any effect on what people are doing. There's a lot of intellectual inertia.
ZIERLER: For the last part of our talk, I'd like to ask some bigger questions about what you've contributed. Do you feel that the case that earthquakes are impossible to predict is somehow counterintuitive?
GELLER: It depends on whose intuition. Going back to 1880, when seismology was starting out, people thought earthquakes were these enormous releasers of energy, and that there had to be some warning that they were coming. That may or may not be reasonable for people to have thought. But anyway, people thought that. Maybe many people still think that. But there are lots of things in science where counterintuitive propositions have turned out to be correct.
ZIERLER: Do you personally believe that at some point in the future, seismology will have the tools and theories to make earthquakes predictable, or that will never happen?
GELLER: The latter, because of the complexity. But as I already said, that's just speculation. I'll let people 100 years from now answer that. All I can say is that all prediction efforts up until now have not only failed, but as we learn more and more about earthquakes, what we learn about is their complexity.
ZIERLER: Have you also found an added layer of difficulty simply because people want to believe that earthquakes are predictable because that's just good policy planning? If we know when earthquakes will happen, we can plan accordingly. Has it been difficult getting beyond what people hope to be the truth?
GELLER: It's a complicated issue because many scientists deliberately have exaggerated the likelihood of their research being able to lead to implementation of earthquake prediction as a way to get funding. Particularly in Japan, but now maybe also in the United States, governments want to allocate funding to useful applied research rather than basic research. There's a kind of feedback that if a guy comes along and says, "This work will contribute to earthquake prediction," they give him a lot of money. Of course, the last thing the people who give him the money want is an evaluation of whether or not their funding allocations were successful. The best thing they can do is double down and keep giving the same guys more money. This seems capable of going on almost indefinitely without the moment of truth ever arriving. When I was younger, I thought this would only last up until the next earthquake, just as the war in Ukraine has revealed all of the problems in the Russian Army that were covered up by the generals and military bureaucrats who were skimming off the money. It's all now painfully obvious what happened, thankfully.
In Japan, we had the Kobe earthquake in 1995, which led to history not being rewritten, but the names of programs being changed to eliminate the term "earthquake prediction" and replace it with "earthquake forecasting" or what have you, but with the same guys in charge. Since the year 2003, these guys were repeatedly making alarmist statements about the "magnitude-9 Nankai Trough great earthquake" that was impending, supposedly. They say there's an "80% probability in the next 30 years." Every year it doesn't happen they just shift the time window by a year. But while repeated alarmist statements were being made about Nankai, the magnitude-9 Tohoku earthquake occurred instead. In its aftermath the government scientists all went on TV and said the Tohoku earthquake was "unforeseeable" ("sotei gai" in Japanese). You would've thought that forecasting a magnitude-9 quake in Nakai that still hasn't happened, while instead an event of a similar size in Tohoku did occur, would've put them out of business. But they actually came out of it with more funding than before. The public and media, even the opposition parties in parliament, no one has demanded to know whether this is real or bullshit. It's very strange. It has to stop at some point, one would think. But I was really wrong in thinking that it would take just one big unpredicted earthquake to blow them away. In addition to the Kobe and Tohoku earthquakes, we also had, in 2016, the Kumamoto earthquake. There were actually two earthquakes in a two-day period, the second being larger, magnitude-7.2. It caused over 200 deaths. That was also in a supposedly very low-risk region. But these repeated failures of the long-term predictions haven't really led to any kind of shakeup. I think that's also true in the US, though.
ZIERLER: The commonalities are there, you're saying.
ZIERLER: For my last question, I'd like to take it back to Caltech. I'm curious if you can reflect on how the approach to science, the way you were educated at Caltech, influenced you throughout your career.
GELLER: I think that, first of all, the influence of Feynman at Caltech was very strong when I was there. He sort of wrote these light, popular books later in his career. But earlier in his career, besides being a leading researcher, he was a great teacher and popularizer of serious science. Reading his textbook  and books like The Character of Physical Law , and hearing his occasional lectures, had a big influence on me. I think Feynman inculcated this philosophy of physics and science into people throughout Caltech. To me, the most important and memorable thing he said was that if your theory doesn't agree with the data from experiments or observations you have to toss it out as being wrong. That's something that no scientist should ever forget.
If I can digress for a moment, in the American Civil War, they trained the soldiers how to fire these very complicated muskets. But on the actual battlefield, with the enemy shooting at them, a lot of soldiers lost their composure. They loaded one musket ball and cartridge into their rifle without firing it, then another and another. When they were shot by the enemy and their rifles were picked up, these huge wads of unfired bullets and cartridges were found in the rifles.
Even today we have a lot of folks looking for earthquake precursors who just make retrospective anecdotal observations. They all studied physics but somehow they've forgotten that they have to have a testable hypothesis, and then test whether or not the alleged precursors have a statistically significant correlation with the earthquakes. So they've allowed themselves to forget what Feynman said, which I'm sure their own professors also taught them, about the essence of physics. We need to stop publishing anecdotal precursor reports and instead do systematic and objective hypothesis testing. If we do that we'll find out in a hurry whether or not there are repeatably observable precursors that are statistically significant. I strongly suspect the answer is negative, but I'd be delighted if we could arrive at a definite answer one way or the other and stop spinning our wheels.
ZIERLER: That's a great point to end on. Bob, I'd like to thank you so much. This has been great fun spending this time with you. I'm so glad we were able to do this. I'd like to thank you so much.
 Tsuboi, C., Wadati, K. & Hagiwara, T., 1962. Prediction of earthquakes: Progress to Date and Plans for Future Development. Earthquake Prediction Research Group, University of Tokyo (English translation). https://doi.org/10.15083/0002002832
 Knopoff, L., Aki, K., Allen, C.R., Rice, J.R. & Sykes, L.R. (eds), 1996. Earthquake Prediction: The Scientific Challenge (Colloquium Proceedings). Proc Nat. Acad Sci.USA 93, 3719–3837.
 Takeuchi, H., Uyeda, S. & Kanamori, H., 1967. Debate About the Earth: Approach to Geophysics through Analysis of Continental Drift. Translated by Kanamori, K., Freeman Cooper, San Francisco.
 Liu, H.-P., Anderson, D.L. & Kanamori, H., 1976. Velocity dispersion due to anelasticity; implications for seismology and mantle composition. Geophys. J. R. Astr. Soc. 47, 41–58.
 Forsyth, D.W. & Uyeda, S., 1975. On the relative importance of the driving forces of plate motion. Geophys. J. R. Astr. Soc. 43, 163–200.
 Whitcomb, J.H., Garmany, J.D. & Anderson, D.L., 1973. Earthquake prediction: Variation of seismic velocities before the San [Fernando] earthquake. Science 180, 632–635.
 Scholz, C.H., Sykes, L.R. & Aggarwaal, Y.P., 1973. Earthquake prediction: a physical basis. Science 181, 803–810.
 Allen, C.R. & Helmberger, D.V., 1973. Search for temporal changes in seismic velocities using large explosions in southern California, in Proc. Conf. Tectonic Problems of the San Andreas Fault System. Eds. Kovach, R.L. & Nur, A., Vol. 13, pp. 436–445, Stanford Univ. Publ. Geol. Sci., Stanford, CA.
 Geller, R.J., 1997. Earthquake prediction: a critical review. Geophys. J. Int. 131, 425–450.
 Kanamori, H. & Anderson, D.L., 1975. Theoretical basis of some empirical relations in seismology. Bull. Seism. Soc. Am. 65 1073–1095.
 Geller, R.J., 1976. Scaling relations for earthquake source parameters and magnitudes. Bull. Seism. Soc. Am. 66, 1501–1523.
 Kanamori, H., 1977. The energy release in great earthquakes. J. Geophys. Res. 82, 2981–2987.
 Kanamori, H., 1970. Synthesis of long-period surface waves and Its application to earthquake source studies—Kurile Islands Earthquake of October 13, 1963. J. Geophys. Res. 75, 5011–5027.
Kanamori, H., 1970. The Alaska earthquake of 1964: Radiation of long-period surface waves and source mechanism. J. Geophys. Res. 75, 5029–5040.
 Kanamori, H., 1977, op. cit.
 Geller, R.J. & Kanamori, H., 1977. Magnitudes of great shallow earthquakes from 1904 to 1952. Bull. Seism. Soc. Am. 67, 587–598.
 Geller, R.J. & Mueller, C.S., 1980. Four similar earthquakes in central California. Geophys. Res. Lett. 7, 821–824.
 Geller, R.J., 1990. Problems of tenure in Japan. Nature 345, 380.
 Geller, R.J., 1991. Shake-up for earthquake prediction. Nature 352, 275– 276.
Geller, R.J., 2011. Shake-up time for Japanese seismology. Nature 472, 407–409.
 Lighthill, J.H., ed., 1996. A Critical review of VAN. World Scientific, Singapore.
 Geller, R.J., 1996. VAN: A Critical evaluation, in A Critical review of VAN, ed. J.H. Lighthill, World Scientific, Singapore, pp. 155–238.
 Geller, R.J., Jackson, D.D., Kagan, Y.Y. & Mulargia, F., 1997. Earthquakes cannot be predicted. Science 275, 1616–1617.
 Debate on VAN (special issue of Geophys. Res. Lett.). May 27, 1996, volume 23, number 11.
 Huizenga, J.R., 1992. Cold Fusion: The Scientific Fiasco of the Century, University of Rochester Press, Rochester, NY.
 Silver, P. & Wakita, H., 1996. A search for earthquake precursors. Science 273, 77–78.
 Evans, R., ed., 1997. Special Section—Assessment of Schemes for Earthquake Prediction. Geophys. J. Int. 131, 413–533.
 Kagan, Y.Y., Jackson, D.D. & Geller, R.J., 2012. Characteristic earthquake model, 1884-2011, RIP. Seismol. Res. Lett. 83, 951–953.
 Stein, S., Geller, R.J. & Liu, M., 2012. Why earthquake hazard maps often fail and what to do about it. Tectonophysics 562-563, 1–25.
 Mulargia, F., Stark, P.B. & Geller, R.J., 2017. Why is probabilistic seismic hazard analysis (PSHA) still used? Phys. Earth Planet. Int. 264, 63–75.
 Feynman, R.P., Leighton, R.B. & Sands, M., 1964. Lectures on Physics, vols 1-3, Addison-Wesley, Palo Alto.
 Feynman, R.P., 1967. The Character of Physical Law, MIT Press, Cambridge, MA.