Scientific Realism

Core Idea

Scientific realism holds that successful scientific theories are approximately true descriptions of reality, including unobservable entities. Electrons, atoms, and fields are real, not merely useful fictions. The 'no miracles' argument contends that scientific success would be miraculous if theories were fictions rather than truth-tracking. However, realism faces challenges: the pessimistic induction (past false theories were also successful), underdetermination (multiple theories fit the data), and apparent conventionality of frameworks suggest empirical success alone doesn't establish truth.

Explainer

The central question of scientific realism is deceptively simple: when science succeeds — when it produces accurate predictions, useful technologies, and coherent explanations — should we believe its theories are *true*? And specifically, should we believe in the entities those theories describe, including ones we cannot directly observe? Scientific realism says yes: electrons, quarks, black holes, and genes are real features of the world, not just bookkeeping tools. The alternative — treating theoretical entities as useful fictions that organize observations without describing anything real — is called anti-realism, and it comes in several flavors, including instrumentalism and constructive empiricism.

The strongest argument for realism is the no-miracles argument: the predictive success of science would be a cosmic coincidence — a miracle — if our best theories were not at least approximately tracking truth. Quantum electrodynamics predicts the magnetic moment of the electron to eleven decimal places. Penicillin works because bacteria really do have cell walls that the drug really does disrupt. If these theories were just mathematical conveniences with no contact with reality, their stunning predictive accuracy would be inexplicable. The best explanation of scientific success is that theories are approximately true and theoretical entities genuinely exist. This is an inference to the best explanation — the same inferential pattern scientists use inside their theories.

The most powerful challenge comes from history. The pessimistic meta-induction, associated with Larry Laudan, observes that the history of science is a graveyard of successful but false theories. Phlogiston theory explained combustion. The luminiferous ether was posited to explain light propagation and generated correct predictions. Caloric theory predicted heat flow correctly for a time. All were eventually abandoned as false. If successful past theories were nevertheless false, shouldn't we expect our currently successful theories to be false too? The realist's reply is typically to distinguish *approximate* truth from exact truth, and to argue that what survives across theory change is often the structurally important part — the parts that do the predictive work.

A second challenge is underdetermination: for any body of evidence, multiple incompatible theories can fit it equally well. If the data cannot force a unique theory, how can empirical success establish which theory is true? The realist must argue either that only one theory is genuinely explanatory (not just empirically equivalent) or that theoretical virtues like simplicity and unification are truth-tracking. This opens deep debates about the relationship between explanatory power and truth. Together, the no-miracles argument and the pessimistic induction define the landscape: realism offers the best story about why science works, while anti-realism is motivated by the humbling history of once-successful theories we now reject. The debate turns on whether you think inference to the best explanation is reliable when it reaches beyond what we can observe.