Part V: Realism vs Anti-Realism

Consider the electron. No one has ever seen an electron. We observe tracks in cloud chambers, dots on screens, current flowing through wires — but these are all effects that we attribute to electrons on the basis of theory. The electron itself is an unobservable, theoretical entity posited to explain a range of observable phenomena.

The scientific realist says: electrons are real. They exist independently of our theories and observations. Our theories about electrons are approximately true. The extraordinary precision of quantum electrodynamics — which predicts the electron’s magnetic moment to better than ten decimal places — would be inexplicable if electrons did not exist.

The constructive empiricist says: we can accept quantum electrodynamics as empirically adequate — it saves all the observable phenomena — without committing ourselves to the literal existence of electrons. Perhaps electrons are real; perhaps they are useful theoretical posits that will eventually be replaced by something entirely different. The history of science gives us reason for caution: phlogiston, caloric, and the luminiferous ether were all once “real” entities that turned out not to exist.

The realism debate has deep roots. In the nineteenth century, Ernst Mach argued that atoms were merely convenient fictions — useful calculational devices with no claim to literal existence. His opponent Ludwig Boltzmann insisted they were real. When Jean Perrin’s experiments on Brownian motion (1908) confirmed the atomic hypothesis, many took this as a vindication of realism. As Perrin himself wrote:

“The atomic theory has triumphed. Its opponents, which until recently were numerous, have been convinced and have abandoned one after the other the sceptical position that was for a long time legitimate and perhaps useful.”

But the victory was not as clean as Perrin suggested. The logical positivists of the Vienna Circle revived anti-realist themes by arguing that the meaning of a statement is exhausted by its method of empirical verification. Theoretical terms like "electron" derive their meaning solely from their connections to observable phenomena. When positivism collapsed in the 1960s, scientific realism enjoyed a resurgence — championed by Hilary Putnam, Richard Boyd, and others.

Then in 1980, Bas van Fraassen’s The Scientific Image reinvigorated anti-realism with a sophisticated new position: constructive empiricism. The debate has continued ever since, generating increasingly refined positions and arguments.

Scientific Realism

The view that our best scientific theories provide approximately true descriptions of both observable and unobservable aspects of the world. Realists typically endorse three commitments:

• Metaphysical: There exists a mind-independent world with a definite structure.
• Semantic: Scientific theories should be interpreted literally — theoretical terms genuinely refer.
• Epistemic: Mature, successful theories are approximately true, and we are justified in believing them.

Constructive Empiricism

Van Fraassen’s position that science aims not at truth but at empirical adequacy — saving the observable phenomena. We can accept a theory as empirically adequate without believing its claims about unobservable entities. Acceptance involves commitment to work within the theory’s framework, but belief is reserved for what the theory says about observables.

Instrumentalism

The older view that theories are not descriptions of reality at all but merely instruments for prediction and calculation. On this view, it is a category mistake to ask whether a theory is "true" — the appropriate question is whether it is useful, accurate, or fruitful. Duhem and Mach defended versions of this position.

Structural Realism

A compromise position arguing that what is preserved through theory change is not the nature of theoretical entities but the structural relations described by scientific theories. Worrall’s epistemic structural realism claims we can know the structure of the unobservable world but not its nature. Ladyman and French’s ontic structural realism goes further, arguing that structure is all there is — there are no "objects" underlying the relations.

The realism debate raises important methodological questions. How should the debate be conducted? What kind of evidence is relevant? Both sides employ inference to the best explanation (IBE) as a central argumentative tool — the realist argues that realism is the best explanation of the success of science, while the anti-realist argues that empirical adequacy provides a better explanation.

Some philosophers have argued for a naturalistic approach: the realism question should be settled by examining the actual history and practice of science, not by a priori philosophical argument. If we look at the historical track record, do successful theories tend to be preserved (supporting realism) or overturned (supporting anti-realism)? This approach has generated detailed case studies of theory change in physics, chemistry, biology, and other sciences, providing much richer empirical data for both sides of the debate.

The No-Miracles Argument

The most influential argument for realism, articulated by Hilary Putnam: “The positive argument for realism is that it is the only philosophy that doesn’t make the success of science a miracle.” If our theories about electrons, DNA, and tectonic plates weren’t at least approximately true, how could they yield such spectacularly accurate predictions and enable such powerful technological applications?

The Pessimistic Meta-Induction

The most devastating argument against realism, pressed by Larry Laudan: the history of science is a “graveyard” of theories that were once empirically successful yet turned out to be fundamentally false. Caloric theory, phlogiston, the luminiferous ether, Newtonian gravitational theory — all were spectacularly successful in their day. If past successful theories were false, what warrant do we have for believing present ones are true?

Closely connected to the realism debate is the problem of underdetermination of theory by evidence. Pierre Duhem and W.V.O. Quine showed that any body of evidence is compatible with multiple, mutually incompatible theories. If the evidence cannot uniquely determine which theory is true, the realist’s claim that we should believe our best theory seems undermined.

Duhem’s version focuses on the fact that scientific hypotheses are never tested in isolation — they are always tested in conjunction with auxiliary assumptions. When an observation conflicts with a prediction, we can always save the hypothesis by modifying an auxiliary assumption instead. This makes it impossible to conclusively falsify a single hypothesis, let alone establish it as uniquely true.

The anti-realist uses underdetermination to argue that even our best theories may have empirically equivalent rivals — alternative theories that make exactly the same observational predictions but posit different unobservable entities and processes. If such rivals exist, we have no empirical grounds for preferring one over another, and belief in the specific unobservable ontology of any one theory is unwarranted. The realist responds that genuinely empirically equivalent rivals are extremely rare in practice, and that theoretical virtues (simplicity, explanatory power, fertility) provide legitimate grounds for preferring one theory over another.

The Ether and Electromagnetic Fields

The luminiferous ether is perhaps the most cited case in the realism debate. Nineteenth-century physicists posited an ether — a mechanical medium — to explain the propagation of light waves. Fresnel’s ether theory was spectacularly successful, predicting the Arago/Poisson bright spot and many optical phenomena with remarkable precision. Yet the ether does not exist; Maxwell’s electromagnetic theory and Einstein’s special relativity showed that light propagates through empty space. The anti-realist cites this as a paradigmatic case of a successful theory with a false ontology. The structural realist counters that Fresnel’s equations survived intact — the structure was preserved.

Atoms: From Fiction to Fact

The atomic hypothesis presents the opposite narrative. Ernst Mach and the energeticists dismissed atoms as useful fictions. Yet through converging lines of evidence — Brownian motion, X-ray crystallography, cathode rays, radioactivity — atoms were established as real entities. Jean Perrin’s work on Brownian motion was particularly persuasive because he used thirteen independent methods to determine Avogadro’s number, all converging on the same value. This convergence of independent methods is one of the realist’s strongest arguments: it would be an extraordinary coincidence if atoms were mere fictions yet all these methods pointed to the same “fictional” number.

The Higgs Boson

The discovery of the Higgs boson at CERN in 2012 provides a contemporary case study. The particle was predicted in 1964 by several independent groups (Higgs, Brout, Englert, Guralnik, Hagen, Kibble) on the basis of theoretical considerations about electroweak symmetry breaking. Its mass was not precisely predicted, but the mechanism was a structural consequence of the Standard Model. The discovery, nearly fifty years after the prediction, is cited by realists as a paradigmatic success of theoretical physics — the theory posited an entity, predicted its properties, and the entity was found. Anti-realists may respond that the discovery confirms the theory’s empirical adequacy but not necessarily the literal truth of the entire theoretical framework.

The Electron Gun

Imagine a physicist using an electron gun to probe the structure of a crystal. She adjusts the beam energy, interprets diffraction patterns, and draws conclusions about the crystal’s lattice spacing. Is she justified in believing electrons exist? Hacking argues yes — she is using electrons as tools, manipulating their properties to investigate other things. You cannot use a fiction to causally interact with reality.

The Martian Scientist

Suppose a Martian scientist independently develops a theory of electromagnetism. She has never communicated with earthling scientists, uses a different notation and different theoretical concepts, yet her theory makes exactly the same predictions as Maxwell’s. The realist argues this convergence would be best explained by both theories being approximately true descriptions of the same reality. The anti-realist might argue it shows only that both theories are empirically adequate.

Van Fraassen’s Mouse

Van Fraassen imagines a mouse whose sensory abilities differ from ours. The mouse can detect certain chemicals but cannot see visible light. For the mouse, molecules are “observable” (via chemical detection) but rainbows are “unobservable.” The constructive empiricist draws the observable/unobservable line at the limits of human perception, not the mouse’s. But why should the epistemically relevant boundary be our boundary? This thought experiment highlights the anthropocentrism of constructive empiricism.

The role of scientific models complicates the realism debate. Models are idealized, simplified representations of reality — they deliberately distort or omit features of the target system. The ideal gas law, the Lotka-Volterra model of predator-prey dynamics, and frictionless planes are all models that are strictly false but scientifically useful.

The anti-realist argues that if science works primarily through models that are known to be false, this undermines the realist’s claim that the aim of science is truth. The realist responds that models can be approximately true in relevant respects even if literally false in others — the distortions are controlled and understood. The debate about models thus recapitulates the broader realism debate at a more fine-grained level.