Nobody has ever seen an electron. Physicists infer electrons from tracks in cloud chambers, readings on meters and the behavior of circuits, and they use them every day to design chips and screens. So what should we say: that electrons exist, or that the electron is a useful fiction that organizes our predictions? This is the debate over scientific realism, and it is one of the liveliest in the field.
Two broad positions
Scientific realism holds that our best theories are approximately true, including what they say about things we cannot observe. Electrons, genes and black holes are as real as tables. Anti-realists deny that we are entitled to that belief. Their oldest form is instrumentalism, which treats theories as tools for prediction rather than literal descriptions of reality. When Copernicus's great book appeared in 1543, an unsigned preface added by the theologian Andreas Osiander told readers that its hypotheses need not be true or even probable, so long as they made the calculations come out right. Copernicus himself believed the Earth moved, but the attitude never died.
The no miracles argument
The realist's favorite argument appeals to success. Theories about unobservable things have predicted phenomena no one had imagined: new elements, new particles, the bending of starlight. In 1975 the American philosopher Hilary Putnam argued, in substance, that realism is the only philosophy that does not make the success of science a miracle. The no miracles argument says the success of science would be an inexplicable coincidence unless its theories were roughly true. If there were no electrons, why would electronics work?
The pessimistic meta-induction
The American philosopher Larry Laudan answered in 1981 with a list. History is crowded with theories that were successful in their day and are now regarded as false: the crystalline spheres of ancient astronomy, the humors of ancient medicine, the phlogiston of eighteenth century chemistry, and the caloric theory, which explained heat as an invisible fluid. If success was no guarantee of truth then, why should it be now? This argument, the pessimistic meta-induction, reasons that because many successful past theories proved false, our current theories probably will too.
Constructive empiricism
The Dutch American philosopher Bas van Fraassen offered a sophisticated anti-realism in 1980. On his constructive empiricism, the aim of science is empirical adequacy: getting right everything a theory says about what can be observed. To accept a theory is to believe that much, and no more. A scientist may use the language of electrons with full commitment as a working practice, while believing only that the theory's observable predictions hold. Van Fraassen also offered a rival to the no miracles argument. Successful theories, he suggested, are like well adapted organisms: they survive because the unsuccessful ones were weeded out, not because they mirror reality.
Critics questioned the line between observable and unobservable. The American philosopher Grover Maxwell had argued in 1962 that we see through a window, through spectacles, through a magnifying glass, a microscope, an electron microscope, with no point at which seeing suddenly becomes inferring. There is no sharp line between observing a thing and detecting it with instruments, so a philosophy that rests great weight on that line looks arbitrary. Van Fraassen accepted that the line is vague but held that vague distinctions can still be real, as with day and night.
Middle positions
Two positions try to keep what is best in each side. The British philosopher John Worrall proposed structural realism in 1989. What survives when theories are replaced, he argued, is not their picture of what things are but the mathematical structure of successful theories, the equations relating one quantity to another. We should be realists about that structure and agnostic about the rest.
The Canadian philosopher Ian Hacking proposed entity realism in 1983. He described a laboratory experiment in which physicists sprayed electrons onto a tiny ball to change its charge, in order to study something else. They did not need to know the final theory of electrons; they needed to know how to use them. Entity realism holds that things we can reliably manipulate as tools are real, whatever our theories eventually say about them.
Worked example: a spot in the shadow
In 1818 the French Academy of Sciences held a competition on the nature of light. The engineer Augustin Fresnel submitted a theory treating light as waves in an all pervading ether. One judge, the mathematician Siméon Denis Poisson, who preferred a particle theory, worked out a strange consequence: if Fresnel was right, a bright spot should appear at the very center of the shadow of a small round disk. That seemed absurd. Another judge, François Arago, tried the experiment and found the spot. Fresnel won the prize. Here is the no miracles argument at its strongest: a novel, unlikely prediction came true. Yet the ether does not exist, as later physics concluded, and so here is the pessimistic meta-induction at its strongest too. Worrall pointed to this very case. Fresnel's equations for how light reflects and refracts passed unchanged into Maxwell's theory of electromagnetism, even though the ether they were meant to describe was abandoned. What Fresnel got right, on this reading, was the structure, and that is what survived.
