Scientific realism is a position in the philosophy of science concerning what science achieves and what scientific theories represent. It holds, in its most common formulation, that mature and successful scientific theories are approximately true descriptions of the world - including entities and processes that are not directly observable - and that the progress of science involves converging toward more accurate representations of reality. Anti-realist positions, by contrast, counsel epistemic caution about unobservables and prioritize accuracy about what can be directly observed or measured. The position stands in contrast to various forms of anti-realism, including instrumentalism (which treats theories as tools for prediction rather than descriptions of reality) and constructive empiricism (which limits scientific knowledge to observable phenomena). Scientific realism is one of the central debates in the philosophy of science and bears on questions in metaphysics, epistemology, and the history of science.
The debate over scientific realism has been active since at least the early twentieth century and remains unresolved. Realists typically appeal to the “no-miracles argument,” which holds that the predictive success of science - including the novel prediction of previously unobserved phenomena - would be inexplicable if theories were not at least approximately true. Anti-realists counter with the “pessimistic meta-induction,” pointing to the historical record of successful but ultimately abandoned theories (e.g., caloric, phlogiston, the luminiferous ether) as evidence that even successful theories may be radically mistaken about unobservable entities.
Intermediate positions have emerged in response to these challenges. Structural realism, developed in different forms by John Worrall and James Ladyman, holds that science accurately captures the relational or structural features of the world even if the nature of underlying entities remains unknown. Entity realism, associated with Ian Hacking and Nancy Cartwright, holds that science successfully identifies real entities when scientists can causally manipulate them, while remaining skeptical of the truth of comprehensive theoretical frameworks. Selective realism attempts to distinguish which parts of successful theories are responsible for their predictive success and claims approximate truth only for those parts.
The debate also intersects with questions about scientific explanation, the nature of laws of nature, and the status of scientific models. Some philosophers argue that the realism debate is underdetermined by evidence - that no empirical finding could decisively settle it - while others hold that it is a substantive metaphysical question with real stakes for how science is understood.
There is no consensus within philosophy of science on scientific realism. It remains one of the discipline's most contested foundational questions. Some surveys of professional philosophers show a plurality inclination toward realism, but significant minorities hold anti-realist or agnostic positions. See the Philosophy of Science Consensus page for survey data and disciplinary distributions.