Reproducibility. What is nature?
There are two key characteristics of science that underlie its immense power but limit its scope: reproducibility and Clarity (which we will discuss in the next post). It was said that whenever the great 19th century scientist Michael Faraday heard of some new phenomenon the first thing he would do was attempt to reproduce the effect in his own laboratory. He explained that his imagination had to be anchored in what he called the "facts", and he understood that science is concerned with reproducible experimental phenomena. For a phenomenon to be a question of science, it has to give reproducible results, independent of who, where, and when of the experiment.
Induction is often touted as the defining philosophical method of natural science, but as a procedure it is hardly more than a formalization of the everyday processes of discovery practiced by humans from their earliest conscious moments. What Francis Bacon did in laying the foundations for the Scientific Revolution was not so much to explain induction as to insist that science had to be practical ("for the relief of man's estate" as he put it). Practical technique demands reproducibility. The knowledge that gives rise to useful technology has to be knowledge that is reproducible and gives rise to tangible effects. These are precisely the characteristics of modern science.
Establishing confidence in reproducible knowledge, certain enough for practical application and meeting our expectations for natural science, requires a deliberate approach to experimentation which is a formalization of the notion of reproducibility. Isaac Newton's famous demonstration that white light is in fact composed of a spectrum of light of different colors is often cited as a "crucial experiment". In his letter to the Royal Society of February 1672 he relates his experiments with the refraction of light through a prism. He talks about various ideas he ruled out as possible explanations of the observations and then says:
Figure 1: Newton's experiment with prisms.
This done, I took the first Prisme in my hand, and turned it to and fro slowly about its Axis, so much as to make the several parts of the Image, cast on the second board, successively pass through the hole in it, that I might observe to what places on the wall the second Prisme would refract them. And I saw by the variation of those places, that the light, tending to that end of the Image, towards which the refraction of the first Prisme was made, did in the second Prisme suffer a Refraction considerably greater then the light tending to the other end.1
In a remarkably short time, the acceptance of this demonstration became practically universal because the key qualitative features, and even the quantitative aspects, could be reproduced at will by experimenters with only a moderate degree of competence.
Nevertheless, an important objection arises in response to the view that science is utterly dependent on this kind of reproducibility: what about observational sciences, like astronomy? The heavenly bodies are far outside our reach. We cannot do experiments on them. Yet who in their right mind would deny astronomy is science? Or consider the early stages of botany or zoology. For centuries, those disciplines consisted largely of systematic gathering, cataloging and classifying of samples of species. Surely it would be pure physicist's arrogance to say that botany or zoology were not, even in their classification stages, science. Are the observational sciences exceptions to the principle that science requires reproducibility?
No. Astronomy was from the pre-industrial age the archetype of reproducibility. It wasn’t just because the heavens showed remarkable systematically repeated cycles that it commanded the attention of so many philosophers. It was because the repeatability gave astronomers the ability to predict with amazing precision the phenomena of the heavens that astronomy appeared almost mystical in its status. The independence of place and observer was satisfied by astronomy with superb accuracy.
There are, of course, unique phenomena in astronomy. On 4th July 1054, astronomers in China first observed a new star in the constellation of Taurus. Its brightness grew visibly day by day. During its three brightest weeks it was reported as visible in daylight, four times brighter than the evening star (the planet we now know as Venus). It remained visible to the naked eye for about two years. We now know these astronomers were observing a supernova. If this were the only supernova ever observed, then we would probably be much more reticent to regard the event with credence. But of course it is not. With modern telescopes, supernovae in other galaxies can be regularly observed.
The SN1054 supernova gave rise to the beautiful Crab Nebula, seen below:
Figure 2: Photograph by the Hubble Telescope of the Crab Nebula.2
Thus, one can thereby get readily accessible reproducible evidence of its date. The expansion rate of the Crab Nebula can be established by comparing photographs separated in time. One can then trace that expansion backward and discover when the now-expanding rim must have been all together in the local explosion. This process, applied by an undergraduate at Dartmouth College to photographs taken 17 years apart, gives us a date in the middle of the 11th century, which lines up perfectly with the recorded observations of the Chinese astronomers. The Crab Nebula's supernova, though having its own unique features, was an event of a type represented by numerous other examples. It was observable to, and recorded by, multiple observers. It left long-lived evidence that for years could be seen by anyone who looked. These are the characteristics of reproducibility in the observational sciences.
One can go through the same exercise for botany, zoology, and geology, all of which have had major phases where they were largely observational. Even so they require multiple repeatable examples of the phenomenon or specimen under consideration. Science does not require that these can be produced at will in the way that a laboratory experiment can. Observations of interest may occur only at certain times (for example eclipses) or in certain places (for example in a specific habitat), over which we might have little or no control. But it does require that multiple examples exist reproducibly.
A second important objection to the assertion that science requires reproducibility concerns the occurrence of random phenomena. If science is the study of the world in so far as it is reproducible, why does probability -- the mathematical embodiment of randomness, the ultimate in non-reproducibility -- play such a prominent role in modern physics? Quantum mechanics can calculate accurately the probability of events but not predict them individually in a reproducible way. Where does that leave the view that science demands reproducibility? Did 20th century physics in fact abandon that principle? No. Instead, science presses up against the limits of reproducibility. The world is not completely predictable even in principle, but science wants to describe it as completely as possible, to the extent that it is. So when up against unpredictability, science invokes deterministic mathematics, but uses the mathematics to govern just the probability of the occurrence of events. Probability is, in a sense, the extent to which random events display reproducibility. Science describes the world in terms of reproducible events to the extent that it can be described that way.
A third challenge to the principle of reproducibility lies in the types of events that are inherently unique. How can reproducibility be a principle applied to the Big Bang origin of the universe? Or how can we apply principles of repeatability to the origin of life on earth, or to the details of how the earth's many species got here? The answer is, for specific unique events of history, evidence based on scientific analysis is rarely decisive, but much of natural history is instead about the broad sweep of development of the universe, the solar system, the planet, or the earth's creatures. In other words, questions of natural history are usually about generalities, not particularities. They are about issues giving rise to repeated observational examples, not single instances. The Big Bang theory of the origin of the universe is a generality confirmed by a multitude of observations that show the same result.
Figure 3: Temperature, atmospheric carbon-dioxide, and dust record from the ice cores from Vostok, Antarctica.3
The same goes for the history of the earth's climate, for example. It is discovered in all sorts of ways, from tree rings to paleontology. Perhaps the most convincing and detailed information comes from various types of "cores" sampled from successively deposited strata (layers of rock and soil) in the Earth. Their results are reproducible. If an ice core is drilled in the same place, the results one gets are the same. Deep-sea-bed sediment cores from all over the globe agree with the ice cores.
Science requires reproducibility. But in many fields of human knowledge the degree of reproducibility we require in science is absent. This absence does not in my view undermine their ability to provide real knowledge. On the contrary, the whole point of my analysis is to assert that non-scientific knowledge is real and essential, just not scientific.
Sociologists today acknowledge that sociology does not offer the kind of reproducibility that is characteristic of the natural sciences. Even so, they feel they must insist on the title of science, because of the scientism of the age.
History is a field in which there is thankfully less science envy. Obviously history, more often than not, is concerned with events in the past that cannot be repeated. History is crucial knowledge but cannot be made into a science.
The study of the law (jurisprudence) is a field whose research and practice that cannot be scientific because it is not concerned with the reproducible. The circumstances of particular events cannot be subjected to repeated tests or to multiple observations.
Economics is a field of high intellectual rigor, but the absence of an opportunity for truly reproducible tests or observations and the impossibility of isolating the different components of economic systems means that economics as a discipline is qualitatively different from science.
Politics is a field, if there ever was one, that is the complete contradiction of what scientists seek in nature. It seems a great pity, and perhaps a sign of the scientism we are discussing in this series, that the academic field of study is referred to these days almost universally as Political Science.
These disciplines do not lend themselves to the epistemological techniques that underlie natural science's reliable models and convincing proofs. They are about more indefinite, intractable, unique, and often more human problems. In short, they are not about nature.