Tuesday, 7 August 2018

The Structure of Scientific Revolutions

That inevitable moment in the life of every pretentious pseudo-intellectual has arrived: I have read Thomas Kuhn’s Structure of Scientific Revolutions, and now I will pontificate about it.

If you have no idea what I am talking about, the blurb tells us that Structure, first published in 1962, was “a landmark event in the history and philosophy of science”. The Introductory Essay in my copy opens: “Great books are rare. This is one. Read it and you will see.” You need more? Structure was the source of the term “paradigm shift”, and is a hell of a lot shorter than the source of the term “Catch 22”. (Admittedly less funny.)

Some preliminaries. I read the book once in 2015, prompted by those it has influenced. From the fans, the sociologist Harry Collins, and from the ranks of the detractors, filmmaker Errol Morris. In particular, Morris wrote a series of New York Times articles about Kuhn, starting with an episode from his time as Kuhn’s graduate student, when Kuhn ended an argument by throwing an ashtray at him. Now Morris has expanded his essays into a book. It is called “The Ashtray”. In preparation for reading it, I decided to read Structure one more time.

(I had an email exchange with Harry Collins about the Morris articles. In the last email Collins concluded that Morris was just out to make a buck on Kuhn’s memory. The email ended, “Hand me the ashtray”. Who ever said philosophy was dry and boring?)

The more important preliminary is that I am not, of course, a historian of science, a sociologist of science, or a philosopher. I know nothing of the book’s ancestors (I cannot comment with a shred of authority on who influenced Kuhn, or how original his ideas were), or its descendants (what influence the book had, what has become of its ideas, or the current critical view of it). I bought it, I read it twice, and here are my thoughts.

I. Scientists, Get Some Structure In Your Life. 

The early sections were revelatory. Kuhn provides his take on how science operates, and my reaction was a repeated stream of, “Yes, yes, that’s exactly how it is!”, punctuated by, “Why has no-one told me this before?” Kuhn explains that most of the time scientists have a collected set of opinions, theories, and methods, which is what he calls a paradigm. Within those theories there is a set of legitimate problems to work on. Working on them uncovers more facts to flesh out the details of the theory, or to refine or expand the theory. Kuhn calls this “normal science”, and he characterises it principally as “puzzle-solving”.

Some scientists may be affronted to have their life’s work likened to a child putting together a jigsaw puzzle, but it sounds spot on to me. It also explains why so many scientists seem strangely disinterested in the mysteries of nature that lie only a few steps beneath their work: the motivation, the challenge, the goal, is not necessarily to explain anything, but to solve a puzzle. That is not all scientists, and certainly not the most visible scientists, who honk on continuously about the deep secrets of the universe (to great public cheering), but it is the bulk of the worker bees, who buzz along from one tricky puzzle to the next.

The message this worker bee took from Kuhn, and the bit that had me wondering why no-one had explained it to me before, is that this is good. There is a romantic notion that “real scientists” invent new theories and make breakthroughs, and everyone else is a second-rate hack, a mere wannabe. Bollocks. Science is about filling in the details of our current understanding of nature. That requires skill, dedication, patience, tenacity, and creativity, and if you are a scientist, it is work you can be deeply proud of. If you have a nagging feeling that ultimately you are a failure until someone names a theory after you, then you need to pick up a copy of Structure.

My own field is a perfect example. In 2015 (a few months after I first read Structure) gravitational waves were detected in the LIGO interferometers. The result was announced to huge fanfare as a “breakthrough”, a “discovery”, and even heralded by some as a “revolution”. It was indeed an incredible scientific result, but it was 100% puzzle-solving Normal Science. A prediction from Einstein’s general theory of relativity was measured. It took 100 years; it took many decades just to work out what the theory predicted, let alone measure it. Some of my colleagues who cling to their childhood conception that real science means revolutionary discoveries, will argue that these observations also tested Einstein’s theory, to identify where it fails. Whatever spins your wheels, guys. So far all of the theory “testing” could just as well be called “validation”. For all we know, it will still be validation when these guys breathe their last disappointed breath. But they should not be disappointed. Comparing refined measurements against a theory, and filling in yet more details, is noble and essential work. Cheer up!

At this point you are probably wondering when I am going to get on to the juicy stuff with the revolutions and the paradigm shifts. If so, calm down, and read again from the start. If you are a scientist, your career is almost guaranteed to begin and end with the Normal Science, so get used to it. Plus, there was another revelation from the Normal Science sections.

The existence of a shared set of theories and opinions (i.e., a shared paradigm) is the true engine of scientific progress. This is what defines a science. The point Structure makes is that the alternative is a series of competing schools of thought — as in history, philosophy, economics, religion, etc., etc. What makes a science different is that there is enough evidence to lock everyone into just one agreed picture of what is going on. More precise data or new observations might lead to a new picture, but most of the time the situation is stable, and that stability is essential. A broad consensus exists for long enough to allow scientists to define a set of problems to solve, and to develop ever more sophisticated apparatus to perform experiments. Scientists often forget how important and how powerful that is. Without the consensus picture, without the paradigm, everyone is searching blindly, arguing over definitions and methods and making very slow progress. You cannot expect to win millions of dollars in funding for telescopes, particle accelerators and huge laboratory complexes, if no-one can agree on what fire is.

Put another way: the romantic picture of science as lone searchers who cook up brilliant new theories is exactly what science is not. A discipline filled with these glorious characters might be very exciting, but it is not a mature scientific discipline, and may not become one for a very long time indeed. Until it does, progress will be slow, and more likely nonexistent. (Yes, there is fun to be had applying this criteria to all the “scientific” fields we hate. Knock yourself out.)

So far, so good. But what happens when the current theory stops working? When inconsistencies show up, or data that does not fit? At first, nothing: usually the current picture can be modified to fit the new evidence, or the evidence turns out to be wrong; either way, the problem is likely to go away. But sometimes the problems mount up, and the current picture starts to break down. There is a crisis. Normal science ends, and there is a period much like the pre-science chaos: lots of ideas, which get increasingly desperate. People are in confusion, and some scientists may start to doubt their vocation: “nothing makes sense any more!” Finally there is a new idea, which does fit all the facts, or at least fits them better than before. People may resist it, but if it ends the nightmare of utter confusion, they will eventually settle on it. That becomes the new picture: the paradigm has shifted. If the new picture sticks, then eventually everyone calms down, and Normal Science resumes. Think the Copernican picture of the Earth orbiting the sun, the dawn of quantum mechanics, and Einstein’s theories of relativity. (Kuhn has many examples from chemistry as well.)

Once again, this rings true, and is much more compelling than the notion that a theory is “falsified” and then rejected until someone comes up with a new theory. Rather, a theory clings on for dear life, until it has no choice but to give in to its successor.

I have no idea if anyone ever really believed the “falsifying” thing. Falsifiability is associated with Karl Popper, but how many scientists have read Popper? Or Kuhn, for that matter? Our conception of what science is, and how it works, is so crude that it may as well have been explained to us in a child’s picture book. It probably was explained in a child’s picture book! “Structure” is good on this as well: scientists do not read scientific history, and that is also fine. The consensus of a shared paradigm means that there is nothing to be gained by reading old science, and in fact it may be worse than a waste of time. Newton’s ideas have been refined and are better presented in an undergraduate textbook than in Newton’s own writings. No-one reads Newton, or Maxwell, or Planck. Studying Einstein’s papers is not going to provide a deep insight into the great man’s thinking: in fact, Einstein’s personal thinking is far less useful than the collective wisdom of all those who have worked since, as condensed into text books and current papers. It is not surprising that Kuhn exudes a whiff of disgust at this — he studied the history of science through close reading of original texts — but the point is valuable: this is another of the defining features of a mature science, and another reason why it works so well. If there is an upheaval in your field when someone uncovers a genius’s long-lost manuscript, then your field is probably not a science.

All right. That’s the good stuff. The Normal Science, the sharp insights into what defines science, the passage through crisis and revolution. We are fine all the way up to the end of Section IX. Then it gets screwy.

II. Beneath A Collapsing Structure.

Kuhn talks about revolutions as a change in world view: scientists now see the world differently. Ok. Fine. We have a new point of view. But this is not strong enough for Kuhn. He struggles to impress on us just how fundamental this change is, almost, but not quite, claiming that the world itself changes. As Kuhn got excited about gestalt shifts (is it a drawing of an old lady, or a young girl?), and the nature of human perception, I became confused over just what kind of book this is.  Is it history — describing what has happened in the past, and identifying patterns in the history of science? Or is it amateur cognitive psychology — explaining how the human mind interacts with the real world and builds up (and changes) its interpretation of it? Or is it sociology — how communities of scientists operate? Or, finally, is it philosophy — a thesis about the very nature of reality and our (in)ability to know it? The book’s fans may claim that it is such a work of genius that it is all of these combined; its detractors as a failed mishmash.

Kuhn repeatedly frames the book as tentative, a “sketch”, an essay, to later be fleshed out into a full book. (Kuhn lived another 34 years, and never wrote that book.) If we take it in that spirit, we can almost overlook the oddities of the last sections. As the insightful signals start to sink into philosophical noise, there are still gems to be had. Even the troublesome notion of incommensurability, that scientists from two paradigms work in different worlds and cannot communicate between them, which is the starting point for the complaints Errol Morris makes — even this is a useful insight. It can indeed be incredibly difficult to communicate with those who see the world through a different framework. The history of science is packed with examples of old codgers who could not accept a new theory. Poincare and Lorentz, who got so close to the revelation of relativity, refused to ever accept it. Possibly they could not accept it precisely because their ideas were close to it. Possibly they were just stubborn. Like Planck (almost) said, science advances one funeral at a time. But this is not true of everyone: people do come to understand new theories, they do work out how to translate between them, and they do come to see the advantage of changing their mind.

In the last section Kuhn asks how science has made so much progress. His answer? Scientists are only human: whatever they have done, they will look back and call it progress. He brings up the standard claim that science approaches truth, and dismisses it as naive: the assortment of impressions we have collected during our haphazard intellectual journey can hardly be called truth. This makes it sound like Kuhn is doing a hatchet job on science, but how can he be, when his knowledge of science history and his appreciation of scientists’ achievements is, from page to page, deep and indisputable? When I got to the end of the book I read the last four sections one more time, and simply could not fathom his conclusions, or entirely clarify what they were.

Why does he say so little about the role of experimental data in both challenging paradigms and in resolving revolutions? In particular, why so little appreciation of the incredible improvement in the range and precision of experiments over the decades and centuries? In the Postscript, written several years later, he refers to the “progress” from Aristotle to Newton to Einstein and notes, “…in some important respects, though by no means all, Einstein’s general theory of relativity is closer to Aristotle’s than either of them is to Newton’s.” Whaaat!? He objects to the modern conclusion that Aristotle’s views were merely mistaken. Sure, Aristotle was very clever, perhaps he was the cleverest human who ever lived — why not? — but how could he be anything other than mistaken without a telescope to see the heavens; without a wristwatch, a precision balance or a micrometer to measure time, weight, and distance; without a pipette, a photographic plate, or a particle accelerator? Jeez, the man didn’t even have a bunsen burner! Of course he was mistaken! Ok, I get the point — the role of human misinterpretation, personalities and peer pressure have been overlooked and under-appreciated in the history of science — but reading Structure, I had a queasy sense of the scales being tilted so far the other way as to topple over entirely.

I take the point that data have to be interpreted, and Structure is good on how the same data may tell us different things before and after a revolution, and in between scientists will argue over what the data are saying, not to mention which data are relevant and correct, and which are irrelevant and flawed. The community may also settle on a conclusion before the data warrant it. (A wonderful set of cautionary examples can be found in The Golem.) That is the short-term picture, and it is invaluable to understanding how scientists operate, as communities of human beings, but it is not enough to explain the long-term success of science. With more refined study, a wrong answer will be found out, whether it takes a year, a decade, or a century.

Paradigms do not hit a crisis for social reasons. We would love nothing more than for the Earth to be at the centre of the universe, for there to be no universal speed limit, and for the intuitions we have built up about the world around us to apply at atomic scales as well. We cannot cling to those ideas, not because society does not allow us, but because they are wrong — demonstrably wrong. Surely this is a fundamental defining feature of science? Human beings love to be right, they love to persuade others they are right, and they love to delude themselves. In art, politics and religion you can win by force and by wit, and unfortunately those will get you a long way in science as well, but eventually you hit the hard wall of reality. Reality is the constraint that makes all the difference.

I am not the first scientist to read this book and wonder why he takes a stance on scientific progress so extreme as to seem deliberately obtuse. To step back from the long arc of the human study of nature, through gravitation, electricity, magnetism, atomic structure, molecules, DNA, the big bang, all of it, and to paint scientific progress as akin to communal back-patting, strikes me as truly perverse. He might argue that that is not what he meant, but if he meant something else, perhaps he could have been a little clearer about it?

Lurking behind all of this is a big question: what is science? Is it a Method, a procedure, a technique that humanity has hit upon, like cuneiform writing, Iambic pentameter, or the headlock? Or is it just another of those complex and pedantic cultural rituals, like Kabuki theatre, Catholic Mass, or cricket? Or is it the inevitable means by which an intelligent species can discover the nature of reality, through a powerful feedback loop of observation, analysis and identification of patterns? The first option is the schoolbook fairy tale. The second is the messy but fascinating world of scientific communities, which are well worth studying, and the proper subject of Structure. The third, which I am of course fond of, is nonetheless a wishful philosophical stance, and a waste of words to debate. I would argue, though, that there is scant evidence in all of human history and experience to make us seriously question it, and it is entirely compatible with the second option. It seems Kuhn thought differently, and his book left itself wide open to be embraced by the anti-authoritarian, anti-technology counter-culture of the sixties, and worse since.

That is a pity. It is a pity for him, because the very real achievements of his book have been obscured by the nonsense bred from its last sections. Most importantly, though, it is a pity for the rest of us, who have to live in a world that mistrusts science. Is it really all Kuhn’s fault? I do not know, but I do know someone who believes that it is. So now it is time to spend a few hours in a room with Errol Morris’s Ashtray.

Next: The Ashtray.

The amazing career of Errol Morris (and what it can teach us all).
Errol Morris and Truth.
Errol Morris's A Brief History of Time.
Errol Morris's Wormwood.


  1. Another physicist here, without much training in philosophy of science, but with enough interest in and influence from it that couldn't help chipping in.
    I write about a few places that I think your criticism of Kuhn might have been a bit too quick.

    1) No better place to start from than "Why does he say so little about the role of experimental data in both challenging paradigms and in resolving revolutions?"
    As you kind of alluded to yourself, part of it is that 'what counts as a reliable piece of experimental data' is determined not in the heavens but within the scientific community/institution. Also, what that "reliable" piece of experimental data is 'really' indicating, is to be determined within the frames of the (rather conservative) discipline/paradigm/community.
    More to the point, according to Kuhn (the way I understood him as least), at the very bottom of science is the community of scientists (or the multi-national institution of science, if you will). The very thesis that logic, mathematics, and experimental data are to lead the way, is no more than something that the community has agreed upon (or the institution has perpetuated) for some time now; what is meant by "logic, mathematics, and experimental data" might change from time to time, but also the fundamental role assigned to them might fade away if the community decides so. Perhaps 5 million years from now, there will still be a multi-national "scientific institution", but its pillars wouldn't be logic and experimental data anymore. [This kind of reminds me of Goldie ...]
    So in short, Kuhn reminds us that you and I think experimental data should have the final say, because the "institution of science" has disciplined us to think that way.

    2) The scientific institution has a complicated dynamics, part of it is social, part of it are the experiments, and there are often many other elements as well. Kuhn reminds us that since it's the community that determines the weight to be given to all these other elements, the community itself (that is, the social aspect) should be the main focus when studying the dynamics of science. Along the same lines, paradigms hit a crisis only when the society/community admits a crisis. (This is how I'd interpret Kuhn on the social aspect of a scientific crisis; I think you had a different interpretation in mind when you wrote "Paradigms do not hit a crisis for social reasons.")

    3) On Reality and scientific "progress", I think Kuhn has views largely influenced by evolutionary theory. Kuhn believes that there's a reality out there [even the "idealist" Kant believed in such a reality through Noumenons], but thinks the scientific institution "progresses" just the way a species might 'evolve'. The species evolves to more "complete" (read 'truer') forms with time, but there's no fixed end-point which is attracting the evolution; it's the complex dynamics of the evolutionary process that determines where the species is going. Of course the species likely gets more and more resilient to 'actual threats', coming from the real world (which you and I and Kuhn and Kant would probably agree has its own "real" elements), but that's really how strong the link of a theory to reality needs to be: strong enough for the theory (and its practitioners) to survive in the greater environment of the institution of science.

  2. 4) Now there's the question "what do you mean by that outside reality?" I think Kuhn knew better than to attack this question square in the face. And I think every serious philosopher would try to avoid that kind of discussion in public nowadays. I'm not sure if Kant ever clarified that he is really a "realist" either. Sometimes the time is not right for certain reasonable-sounding questions in a discipline. "Is there an objective reality out there?" might sound benign to you and me, but to publicly address it with philosophical rigour was probably too costly for Kuhn. So if he has, as you say, never clarified his position on this, I wouldn't accuse him of intellectual dishonesty for that.

    5) Finally, and most importantly on the question of truth, there is the anti-positivism legacy of Kuhn.
    'Even in science things are not on that solid a ground', Kuhn reminded us. 'So stop thinking that your own beliefs are truest, and others' less true.'
    I think there is timeless insight in this.
    Especially when certain malignant forms of positivism influence cultural views, they can help rationalize Fascism.

    I'm not saying that "Kuhn was right!" (That'd be stupid; everyone has been wrong in some ways.) But his insights are often underestimated.
    As much as "alternative facts" are a serious problem nowadays in the West, so are lack of empathy (especially cultural) , and Fascism; Kuhn's anti-positivist legacy, one might hope, could serve as an antidote to the latter two.

  3. I'll see if I can address these.

    1&2. I feel like this 'what counts as a reliable piece of experimental data' question is overemphasised in these sorts of discussions. When the blackbody curve was observed (which happened because of improvements in technology), I'm sure people initially doubted the result, since it looked different to what they expected/wanted, but then they kept trying to get the result they wanted, and could not. Planck provided the quantum explanation -- in desperation! -- and since no-one could find a better one, they eventually had to accept it. Can't we separate whatever sociology was going on between all these scientists (which I'm sure was very interesting) with the simple fact that new physical phenomena were discovered and explained?

    3. I think Kuhn's analogy with evolution muddies things. I can accept that science progresses in some random walk, and it's not being drawn "towards" something, but that's quite different from saying that there isn't an objective reality that we describe better and better as we randomly prod away at nature. I guess my real objection would be: sure, you can look at it that way, but why should I agree that it is better than any other analogy?

    4&5. Kuhn could have made all of his insights about how science works without *any* of the philosophy. Why wade into that stuff? That was why I asked the question of just what kind of book Structure is. If he couldn't make a solid argument for some philosophical position, then he should have left it out. Doing science is as good a way as any to learn the dangers of being dogmatic about a particular viewpoint; plenty of scientists understand (and understood) that intellectual openness, empathy, etc., are important, without needing Kuhn to tell them.

    As I said in the piece, there's lots we can take from Structure. I feel that the insights that have been underestimated are those earlier in the book. I tried to make a coherent picture of the later stuff, but just could not. As I said in the Ashtray piece, it was refreshing to stop trying: the notion that the Ptolemaic view of the solar system is just as good as the Copernican one, only different, starts to sound very attractive as you read through Structure (it sounds very kind and fair, that everyone's view is valid), but in the end you just have to admit: no, it was wrong! Trying to swallow that line about Aristotle and Einstein was the last straw.

    1. Re 1&2: if the black body radiation story indeed was as clean as you claim, that'd be an example where the social dynamics has been (luckily!) very efficient. Now, trying to see things from a Kuhnian viewpoint I'd argue that it was the scientific authority of the experimentalists that made the experimental results look valid to the community, and it was the authority of theorists like Planck that made the results seem important; of course the precision of the experimentalists' devices (as confirmed by other experimental authorities) helped to establish their authority, and the prior achievements of Planck (as recognized by other theoretical authorities) helped to establish his. At the end of the day, what happened was that some scientists convinced other scientists of something; in such instances, the internal procedures of the institution of science generates and propagates the story that "new physical phenomena were discovered and explained".
      [Let me be clear, I agree that 'real progress was made', but I don't think we fully understand what that really means. Kuhn partially clarifies the meaning of that sentence by tracking how that sentence is generated by a community of people with their own social dynamics (which he partially addresses).]

      Re 3: that's one of the best analogies that I've seen when it gets to science vs truth. That's Kuhn's insight; that's one perspective; I agree that it's good to have complementary perspectives, and try to improve (or synthesize) all of them too.

      Re 4&5: I suppose Kuhn did his best given the circumstances, and probably thought the philosophical stuff was worth getting noticed even though they weren't as neatly worked out as typical scientists would like. Let's keep in mind that Kuhn's main audience weren't scientists, but philosophers.

      Re Aristotle vs Einstein: I don't know how best to clarify that comment, but I suppose if it's taken with a pinch of salt it might help to clarify somethings for some readers of Structure.

    2. I don't want to get dogmatic about this. There's a lot of value in Kuhn's book. But also a lot that takes quite a few pinches of salt and several cups of generosity to interpret as logically consistent. It feels a bit like he's trying to convince us of his picture of science by writing a book that fits exactly that picture: what it means is entirely a construct of the reader! (I hope it's obvious that I don't mean that entirely seriously.)

      I read your reading of the Planck story as saying that the culture and big cheeses of the time made people take a certain result seriously. I would add, however, that it being a correct result was independent of culture. Sure, others with less respect might have found the same results and been ignored, and no progress made for 50 years. Or if Planck's quantum idea had been wrong, more detailed experiments would have uncovered that, and the idea would have faded away. The order in which we find things out is determined by culture and history and luck and so on, and the way we present it and the language we use, etc., but in the end there are still real things to find and there are still right answers and wrong answers. I'm not sure that's the message most people get from Structure.

      Interestingly, I believe that the history of quantum mechanics and its philosophical issues are discussed in a recent book, "What is real?", which is reviewed alongside The Ashtray in this article. The review was interesting, although I have not read the book.


[Note: comments do not seem to work from Facebook.]