Reflections. I. How do ideas change in science?

Looked at over a long historical stretch, science looks like a pretty substantial edifice. Everyone would agree that we know and understand far more about the nature of the natural world in 2025 than we did in 1800 or 1900 or even 1950. Science is an impressive thing. Yet, looked at up close, it becomes clear that the metaphor of the edifice is poor; it does not capture the dynamism of the process. Science is more like a big, growing construction site with near constant demolition and then rebuilding, rather than steady growth. The foremen are in near-constant disputes about what needs to be done next and how best to do it. This sometimes includes angry exchanges about whether the foundations are sound! From this perspective, the progress of science seems like a miracle. (However, if there is one thing that scientists agree on it' is that miracles cannot be invoked.)

These thoughts were prompted recently by a friend of mine sending me some draft chapters from a book he is writing on what one can and cannot expect from artificial intelligence (AI), in light of what we know about how actual brains function. Obviously, I will not discuss his material but will mention a reflection that this reading prompted. It concerns the confluence of two streams of thinking in the early 1950s. They were incredibly valuable to biology for decades but may now be something of a roadblock, hindering further understanding.

The first was information theory, as first put forward by a mathematician and engineer, Claude Shannon, in a landmark paper published in 1948. This work was a major advance in thinking about how large amounts of information can be encoded, namely in a digital form (sequences of 1s and 0s). It both complemented and extended the work on computers then being pioneered by various leading thinkers, such as Norbert Wiener, John van Neumann and Alan Turing, The idea of encoded information soon played a big role in biology in understanding how DNA carries hereditary information (see Why is the genetic code the way it is?).

The second stream of new thinking was in neurobiology and concerned how electrical messages are transmitted between neurons. This involved work on the large giant axons (projections) of the neurons of squids. It was carried out by Andrew Huxley and Alan Hodgkin, two British biologists who were later honored with the Nobel Prize for their work.¹

These two currents of thinking soon came together in further thinking about how the human brain works, a major problem then and now. The brain, after all, is a biological device for handling huge amounts of information, which it sorts, stores, and uses to produce conclusions that often stimulate actions. It does so via huge numbers of neurons. Might the brain not be the biological equivalent of a very sophisticated computer and one that uses digital information to do its “calculations”?

This has been a powerful idea from the 1950s onwards. Today, it is central to the thinking about AI and is therefore to the AI industry. What are the consequences, however, if the brain is so different in its actual operations from computers that the analogy or simile begins to fail? This is still a minority opinion, certainly in the AI field, but it has attracted some notable thinkers since at least the 1980s.² If this heterodox thought – that the brain is not a computer – were to become the consensus, this would not destroy AI but it would require a new conceptual foundation for this industry (and possibly a new name).

This possibility, however, leads to other thoughts. Is it possible that a highly popular and persuasive idea in science actually discourages people from pursuing other explanations that might have a lot to contribute? In the case of the brain-as-digital-computer, I do not know of any researchers who rejected it and whose research suffered as a result. In general, however, people pursuing ideas outside the mainstream often have extra battles to fight. These involve struggles for funds, recognition, and opportunities for research.

This phenomenon of resistance to new ideas in science was discussed by a philosopher of science, Thomas Kuhn, in 1962. He did so in his book The Structure of Scientific Revolutions. He had originally been a student of physics and his examples were all drawn from the history of physics but he clearly believed that his basic idea applied to all the natural sciences. The basic concept was that at any time, a field’s theoretical structure is governed by a dominant idea. His term for such an idea was “paradigm”; we can think of it as a conceptual framework. Most thinkers and researchers in any field are guided by its reigning paradigm in their scientific work. That work, which Kuhn termed “normal science”, is intended to extend the reach and application of the idea. Normal science often produces new findings and successful papers, growing recognition, easier access to funds.

But what if the reigning paradigm does not explain all the new results that it should? That creates a problem. Furthermore, if the paradigm is actually flawed, these exceptions start piling up. If this happens, the field is in a crisis, which can only be relieved by the advent of a new paradigm, if the latter is able to explain the results the old one could not. When such paradigm replacement occurs, this is what Kuhn called a “scientific revolution”.

Kuhn’s idea was provocative and exciting. When I was in my last year of college in 1964-65, I took a philosophy of science course when Kuhn’s book was only two years old, and I remember the excitement about it. Before Kuhn, philosophy of science was a somewhat dry enterprise concerned with proof and disproof. Kuhn’s work introduced psychology, sociology, history, drama. It was fun! Of course, it was also controversial. One question was how exactly one defined “paradigm”. One scholar in the early days listed 22 variant, though often related, meanings! The question was non-trivial if science was to be seen as a succession of paradigms in each field. Perhaps more important, however, was the question: is Kuhn’s scheme really the way science works? Kuhn’s examples from physics were the overthrow of the Ptolemaic conception of the universe by Copernicus and Kepler and the demotion of Newtonian physics by Einsteinian relativity; his discussion of these was convincing. Did Kuhn’s idea, however, apply more broadly, in chemistry and biology? If so, did old paradigms act as a brake on new paradigms coming to the fore?

Figuratively speaking, a lot of ink has been spilled on these questions without real resolution. I made my own miniscule contribution to the discussions with a short paper arguing that in biology, the big ideas that we subscribe to had filled a vacuum, not overthrown a pre-existing paradigm.³ Needless to say, there was some fierce opposition to this claim.

It depends in part on how one defines “paradigm” and whether earlier ideas, now seen as defective, should be demoted retrospectively from previous paradigm status. Another factor concerns the relative speed of replacement of the old idea. Does it have to be fast to count as a revolution? Also, one may ask what proportion of the people in the field must subscribe to the new idea before it is accepted, who they are, and how strong their resistance is. Long before Kuhn published his book, Max Planck, the great physicist, had opined that for new ideas in science to be accepted, the older scientists, who were often quite attached to the older ideas, had to die off!

Is there one pattern of theory replacement in biology, either a match to the Kuhnian idea or something different? My judgment today is that it is a mixed picture. Although new ideas replace others – that is the essence of science, making the language of paradigm replacement appropriate – many shifts in viewpoint do not happen with the speed and drama that the Copernican and Einsteinian scientific revolutions showed. Where the change is slower, that almost always reflects resistance from scientists, some of whom support the earlier ideas, others of whom simply think that the new idea is not good enough. Furthermore, in some cases, the situation conforms to the picture I sketched many years ago, where a new idea with great explanatory power steps in to fill a previous void.

Take the biggest change of ideas in biology, Darwinian evolution. Though there had been a couple of earlier suggestions that evolution can take place along roughly Darwinian lines, Darwin’s work was a thunderclap, not just filling out the idea but offering much supporting evidence. After Darwin had presented his case for the reality of evolution, there was no turning back amongst the great majority of scientists to the ideas of Biblical creation to explain the diversity of life on Earth. If any change in biology deserves the term “revolution”, this was it.

This was not, however, a revolution in the Kuhnian sense, for two reasons. First, there was no preceding scientific idea about evolution that it replaced, only a religious belief. Hence, one cannot speak of paradigm replacement in this case. Second, there was nowhere near as universal assent to Darwin’s proposed mechanism of evolution, namely the idea of natural selection. Indeed, for reasons both good and bad this idea seemed to die a slow death in the remaining decades of the 19^th century.⁴ Darwin was only vindicated on this major point – the one he cared most about – in the 1930s and ‘40s. This was hardly the pattern of scientific change described by Kuhn.

On the other hand, the ideas that comprised the core of the new science of molecular biology – what genes are chemically and how they “encode” proteins – were established with comparatively great speed, over only a decade (1953-1963) (with respect to the genetic code, again see Why is the genetic code the way it is?). The ideas were revolutionary and fast. However, like Darwin’s evidence for evolution, there was no earlier paradigm that was replaced. Hence, this too was not a classic Kuhnian scientific revolution.

Indeed, part of this great intellectual revolution came undone in the 1970s. A key supposition of the molecular biology revolution was that genes are linear sequences of DNA nucleotides that encode matching linear sequences of matching polypeptide sequences. This “co-linearity” of gene and protein sequence was virtually a sacred tenet of molecular biology in the 1960s. Hence, the discovery in the 1970s that, in eukaryotes, not all of the gene’s sequence was retained in the final RNA copy and translated into proteins, came as a great shock. This was the “genes-in-pieces” idea of 1977-1978. Yet, the other elements of the great molecular biology revolution remained more or less intact (though of course becoming far more complicated in their details). Hence, one might ask: was this a case of Paradigm Lost or Paradigm Modified. (It depends on how one defines “paradigm”, which, as we have noted, is a non-trivial issue.)

However, some changes in biology are radical, relatively fast, and lead to paradigm replacement, along very Kuhnian lines. The ideas from Peter Mitchell on how mitochondria capture energy from food materials to produce ATP provide a sterling example.⁵ There was tremendous resistance to Mitchell’s ideas when he first produced them and he had to fall back on his own resources to establish them. However, once he did so and produced evidence, acceptance was swift. This was an undoubted Kuhnian scientific revolution, which I had overlooked in my 1996 article, as was pointed out to me almost immediately.

Is there any useful conclusion that can be drawn from this admittedly short and incomplete look? Perhaps the only safe conclusion is that there is a great variety of patterns in how ideas evolve and rise or fall in biology. There is no one-size-fits-all explanation. Some new ideas win adherents quickly, others have to wage long fights before they win acceptance. It all depends on the context: whether there are potent competing ideas or not, who is championing the new ideas or fighting them and how well they do in either case, plus other variables of the field and how it is doing at the time. That new unproved ideas should meet resistance, by the way, is not a bad thing. Skepticism is a major value in science, right up there with intellectual honesty and trying to be as accurate as possible.

To return to our starting point: what about the (probably) failing idea of the brain-as-computer? My feeling is that this field is in the crisis stage. The old paradigm is in trouble but a fleshed out alternative is still not there. We will return to this problem in a later article.

1. By a nice coincidence for this article, Huxley and Hodgkin received their Nobel Prize in the same year, 1963, as Watson and Crick did for their groundbreaking work.

2. Two of them are Giulio Tononi, a neuroscientist and consciousness researcher at the University of Wisconsin-Madison and Gerald Edelman (1929-2014), a Nobel Prize winner for his work in molecular immunology but who became a leading neuroscientist.

3. See Wilkins, A.S. (1996). Are there “Kuhnian” revolutions in biology? BioEssays 18(9): 695-696.

4. Bowler, P.J. (1983). The Eclipse of Darwinism: anti-Darwinian evolution theories in the decades around 1900. Baltimore: Johns Hopkins University Press.

5. See Mitchell, P. (1977). Vectorial chemiosmotic processes. Annu. Rev. Biochem. 46:996-1005. Doi/10.1146/annurev.bi. 46.07177.005024.