weofodthignen

Posted by Andrew

https://statmodeling.stat.columbia.edu/2026/02/08/review-of-string-theory-book-from-2004/

https://statmodeling.stat.columbia.edu/?p=51662

I don’t agree with everything that Freeman Dyson writes, but this review of a book by Columbia physicist Brian Greene was pretty good. It’s on the topic of string theory, which has come up before on this blog.

Dyson begins his review on a judiciously positive note:

I recommend Greene’s book to any nonexpert reader who wants an up-to-date account of theoretical physics, written in colloquial language that anyone can understand. For the nonexpert reader, my doubts and hesitations are unimportant. It is not important whether Greene’s picture of the universe will turn out to be technically accurate. . . . Even if many of the details later turn out to be wrong, the picture is a big step toward understanding. Progress in science is often built on wrong theories that are later corrected. . . . Greene’s book explains to the nonexpert reader two essential themes of modern science. First it describes the historical path of observation and theory that led from Newton and Galileo in the seventeenth century to Einstein and Stephen Hawking in the twentieth. Then it shows us the style of thinking that led beyond Einstein and Hawking to the fashionable theories of today. The history and the style of thinking are authentic, whether or not the fashionable theories are here to stay.

After quoting from Greene’s description of string theory, Dyson continues:

This is a fine beginning for a theory of the universe, and maybe it is true. To be useful, a scientific theory does not need to be true, but it needs to be testable. My doubts about string theory arise from the fact that it is not at present testable.

I guess not much has changed in the past twenty-one years!

More interesting than the string-bashing or string-skepticism is Dyson’s age-period-cohort take on the sociology of the cutting edge of physics:

In the history of science there is always a tension between revolutionaries and conservatives, between those who build grand castles in the air and those who prefer to lay one brick at a time on solid ground. The normal state of tension is between young revolutionaries and old conservatives. This is the way it is now, and the way it was eighty years ago when the quantum revolution happened. I [Dyson in 2004] am a typical old conservative, out of touch with the new ideas and surrounded by young string theorists whose conversation I do not pretend to understand. In the 1920s, the golden age of quantum theory, the young revolutionaries were Werner Heisenberg and Paul Dirac, making their great discoveries at the age of twenty-five, and the old conservative was Ernest Rutherford . . . a great scientist, left behind by the revolution that he had helped to bring about. That is the normal state of affairs.

Fifty years ago, when I was considerably younger than Greene is now, things were different. The normal state of affairs was inverted. At that time, in the late 1940s and early 1950s, the revolutionaries were old and the conservatives were young. The old revolutionaries were Albert Einstein, Dirac, Heisenberg, Max Born, and Erwin Schrödinger. Every one of them had a crazy theory that he thought would be the key to understanding everything. Einstein had his unified field theory, Heisenberg had his fundamental length theory, Born had a new version of quantum theory that he called reciprocity, Schrödinger had a new version of Einstein’s unified field theory that he called the Final Affine Field Laws, and Dirac had a weird version of quantum theory in which every state had probability either plus two or minus two. . . . Each of the five old men believed that physics needed another revolution as profound as the quantum revolution that they had led twenty-five years earlier. Each of them believed that his pet idea was the crucial first step along a road that would lead to the next big breakthrough.

Young people like me saw all these famous old men making fools of themselves, and so we became conservatives. The chief young players then were Julian Schwinger and Richard Feynman in America and Sin-Itiro Tomonaga in Japan. Anyone who knew Feynman might be surprised to hear him labeled a conservative, but the label is accurate. Feynman’s style was ebullient and wonderfully original, but the substance of his science was conservative. He and Schwinger and Tomonaga understood that the physics they had inherited from the quantum revolution was pretty good. The physical ideas were basically correct. They did not need to start another revolution. They only needed to take the existing physical theories and clean up the details. I helped them with the later stages of the cleanup. The result of our efforts was the modern theory of quantum electrodynamics, the theory that accurately describes the way atoms and radiation behave.

This theory was a triumph of conservatism. We took the theories that Dirac and Heisenberg had invented in the 1920s, and changed as little as possible to make the theories self-consistent and user-friendly. Nature smiled on our efforts. When new experiments were done to test the theory, the results agreed with the theory to eleven decimal places. . . .

This is fascinating. I’d never thought of the history of twentieth-century physics this way, and it leaves me with some thoughts:

1. The age-period-cohort nonidentifiability, something we’ve seen before when studying public opinion. Dyson is talking about the scientific views of leading physicists, but it’s the same general thing, that you can explain the observed data in multiple ways.

One story is that attitudes of the young were driven by the logic of events: in the 1910s-20s, the foundations of statistics were in a mess and so the young physicists were radical, recognizing the need for revolutionary science; in the 1940s-50s, the foundations were strong and much progress could be made using normal science, hence the young physicists were conservative; in the 1980s-90s, the advances of conventional methods in fundamental physics had trickled to a halt, hence the young physicists were motivated to be radical. Another logic-of-events thing that Dyson could’ve mentioned, but didn’t, is that the 1940s-50s were special in that a huge amount of effort in theoretical physics was going into the design of atomic bombs: for that, what was needed was creativity in the application of existing fundamental theories, not a fundamental restructuring. After the 1950s, the military remained a major funder of physics, but bomb design was no longer the cutting edge.

A different explanation is based on cohorts. The cohort of Einstein, Dirac, etc., achieved success with new fundamental theories and so they kept wanting to do that–they remained radicals all their lives. Reacting against this, the cohort of Feynman, Dyson, etc., achieved success while working within existing theories, so they remained conservatives all their lives. The cohort of Greene, etc., reacted against their fathers and became radicals.

These two stories can coexist; that’s part of the nonidentifiability.

2. Similar things have gone on in statistics. From the 1950s-1980s there was a conservative movement within academic statistics featuring opposition to Bayesian methods (see discussion here), with some of this attitude lingering even into the 1990s (as discussed here). My generation was more radical, favoring developments in many different directions. In recent decades I and others have become more aware of misuses of statistics but we have not framed this as an anti-modeling stance. My point here is not that I’m right and these other people are wrong, but just that, as with Dyson, I see generational differences. Indeed, I’ve talked with some young statisticians who express to me what seem like naive old-fogey attitudes demanding statistical guarantees. Again, these are legitimate differences of opinion, in the same way that it’s perfectly fine for Dyson and Greene to differ on the value of string theory. It’s just interesting to see these sorts of age-period-cohort effects in these contexts.

https://statmodeling.stat.columbia.edu/2026/02/08/review-of-string-theory-book-from-2004/

https://statmodeling.stat.columbia.edu/?p=51662