The demarcation problem

The demarcation problem is the philosophical question of what distinguishes science from non-science, pseudoscience, and other forms of inquiry.⁸ Although the question can be traced to antiquity, its modern formulation derives from Karl Popper’s work in the 1930s, and it remains one of the central problems in the philosophy of science. The question is not merely academic: the legal status of creationism and intelligent design in public education, the regulation of medical treatments, and the authority of scientific testimony in policy debates all depend, implicitly or explicitly, on the ability to distinguish science from its imitations.⁹

Popper’s falsifiability criterion

Karl Popper proposed the most influential single criterion for demarcation in Logik der Forschung (1934, published in English as The Logic of Scientific Discovery in 1959). Popper argued that the defining characteristic of a scientific theory is not that it can be verified — since universal statements can never be conclusively verified by a finite number of observations — but that it can in principle be falsified. A genuinely scientific theory makes predictions that, if they fail, would require the theory to be revised or abandoned. A theory that is compatible with every possible observation, that cannot specify any outcome that would count against it, is not scientific.¹

Popper developed this criterion partly in response to what he perceived as the unfalsifiable character of Marxist historical theory, Freudian psychoanalysis, and Adlerian individual psychology, which he contrasted with Einstein’s general relativity — a theory that made precise, risky predictions (such as the bending of light around the sun) that could have been, but were not, refuted by observation.² The contrast was between theories that exposed themselves to potential refutation and theories that could accommodate any outcome through post-hoc reinterpretation.

Falsifiability has been widely adopted as a working criterion, but it faces well-known difficulties. The Duhem–Quine thesis holds that no individual hypothesis can be tested in isolation; any empirical test involves a network of auxiliary assumptions (about instruments, background conditions, and other theories), and a failed prediction can always be attributed to a failure in the auxiliaries rather than in the core hypothesis.⁸ In practice, scientists routinely protect promising theories from apparent falsification by modifying auxiliary hypotheses — a procedure that Popper himself acknowledged but regarded as permissible only under certain conditions.¹

Kuhn and paradigm structure

Thomas Kuhn’s The Structure of Scientific Revolutions (1962) offered a different approach to demarcation, one grounded in the sociology and history of science rather than in formal logic. Kuhn argued that mature sciences are characterised by the existence of a paradigm — a shared framework of theories, methods, standards, and exemplary problem-solutions that defines a research community’s work. Normal science consists of “puzzle-solving” within the paradigm: extending the framework to new phenomena, refining measurements, and resolving anomalies. A field that lacks a shared paradigm, or that has never developed a puzzle-solving tradition, is in a “pre-paradigmatic” state and does not function as a mature science.³

Kuhn’s framework shifts the question from the logical properties of individual theories to the social and methodological characteristics of research communities. A scientific discipline, on Kuhn’s account, is recognisable by its shared commitments, its ability to generate and solve new puzzles, and its capacity for revolutionary change when the existing paradigm accumulates intractable anomalies. Pseudoscientific fields, by contrast, tend to lack a shared paradigm, to fail at generating new puzzles, and to remain static over extended periods.³

Lakatos’s research programmes

Imre Lakatos sought to mediate between Popper’s emphasis on falsification and Kuhn’s historical approach. In his methodology of scientific research programmes, Lakatos argued that the unit of appraisal is not an individual theory but a series of theories linked by a shared “hard core” of fundamental assumptions. A research programme is “progressive” if it successfully predicts novel facts — if its successive theoretical modifications lead to the discovery of new phenomena that the previous version did not anticipate. A programme is “degenerating” if its modifications are purely defensive, designed only to accommodate anomalies after the fact without generating new predictions.⁴

Lakatos’s criterion has the advantage of permitting temporary failures without immediately condemning a programme as unscientific: a theory that is currently struggling with anomalies may be in a temporary downturn before making a progressive recovery. The criterion condemns a programme only when its pattern of modification is consistently degenerating — when it does nothing but explain away problems without ever predicting anything new. This framework has proven particularly useful in evaluating claims about intelligent design and creationism, both of which critics characterise as degenerating programmes.^{4, 14}

Laudan’s critique

The philosopher Larry Laudan mounted the most influential attack on the demarcation project itself. In his 1983 essay “The Demise of the Demarcation Problem,” Laudan argued that every proposed criterion for distinguishing science from non-science either excludes recognised sciences or includes recognised pseudosciences. Falsifiability excludes theoretical claims in string theory and cosmology that may not be directly testable; puzzle-solving traditions exist in astrology and alchemy; progressiveness is a matter of degree rather than a sharp boundary. Laudan concluded that the search for necessary and sufficient conditions for “science” is a failed philosophical project, and that the more productive question is whether a given claim is well-supported by evidence and sound reasoning, regardless of whether it bears the label “science.”⁵

Laudan’s critique had immediate practical consequences. He argued that the ruling in McLean v. Arkansas (1982), which had relied on demarcation criteria proposed by philosopher Michael Ruse to declare creation science non-scientific, was philosophically unsound — not because creation science is good science, but because the criteria used to exclude it were themselves indefensible.⁷ Laudan’s position was that creation science should be rejected because it is bad science — poorly supported, contradicted by evidence, and methodologically deficient — not because it fails to meet some abstract criterion of “scientificness.”

Despite Laudan’s critique, most philosophers of science have not abandoned the demarcation project entirely. Sven Ove Hansson and Massimo Pigliucci, among others, have argued that the failure to find a single sharp criterion does not mean that no meaningful distinction exists between science and pseudoscience, any more than the difficulty of defining “heap” precisely means there is no difference between a heap and a single grain of sand. The demarcation may be a matter of multiple criteria, family resemblance, or degree rather than a sharp binary — but it is nonetheless real and consequential.^{8, 13}

Demarcation and intelligent design

The creation–evolution controversy has been the most prominent arena in which demarcation criteria have been applied in practice. In McLean v. Arkansas Board of Education (1982), Judge William Overton ruled that creation science fails to qualify as science because it invokes supernatural causation, is not testable, and does not generate predictions — criteria drawn from Popper’s falsifiability and Kuhn’s paradigm structure.¹⁵

Two decades later, in Kitzmiller v. Dover Area School District (2005), Judge John E. Jones III applied a similar analysis to intelligent design. Jones ruled that ID fails as science on multiple demarcation criteria: it invokes supernatural causation, it is not testable or falsifiable, it has not generated peer-reviewed research, and it has not been accepted by the scientific community.⁶ The specific claims of ID proponents were evaluated against each criterion. Michael Behe’s “irreducible complexity” was found to be either falsifiable (and falsified, in the case of the immune system and the bacterial flagellum, for which evolutionary precursors have been identified) or unfalsifiable (if defined in a way that precludes any possible evolutionary explanation).^{6, 10} William Dembski’s “specified complexity” was found to lack a rigorous operational definition and to have generated no empirical research programme.^{11, 6}

Evaluated under Lakatos’s framework, intelligent design exhibits the hallmarks of a degenerating research programme: its theoretical modifications have been uniformly defensive (redefining key terms, shifting from “creation science” to “intelligent design” to “teach the controversy”), and it has not produced a single successful novel prediction. Evolutionary biology, by contrast, continues to generate progressive problem shifts — predicting the locations of transitional fossils, the patterns of molecular phylogenies, and the distribution of endogenous retroviruses — and to solve new puzzles on a regular basis.^{14, 4}

Significance and qualifications

The demarcation problem carries an important qualification. To say that a claim or field is “not science” is not to say that it is worthless, meaningless, or necessarily false. Philosophy, ethics, literary criticism, and theology are not sciences, and none is diminished by that observation. The demarcation problem matters not because non-science is valueless but because different types of claims require different standards of evaluation, and because the authority of science in education, medicine, and public policy depends on the ability to distinguish empirically grounded claims from those that merely adopt the trappings of empirical inquiry.^{8, 9}

Paul Feyerabend’s radical critique in Against Method (1975) pushed this point further, arguing that no fixed set of methodological rules governs all successful science, and that the rigid enforcement of demarcation criteria can itself be an obstacle to scientific progress.¹² Feyerabend’s “epistemological anarchism” remains a minority position, but it serves as a useful caution against treating demarcation criteria as infallible diagnostic tools rather than as fallible but useful guides to the epistemic quality of inquiry.

The demarcation problem thus remains open as a philosophical question while remaining indispensable as a practical one. The criteria proposed by Popper, Kuhn, and Lakatos — testability, paradigm structure, and progressiveness — do not individually provide a necessary and sufficient definition of science, but taken together they identify a cluster of properties that reliably distinguish the most productive forms of empirical inquiry from their imitations. The creation–evolution controversy, and its legal history in particular, demonstrates that the stakes of this philosophical question extend far beyond the seminar room.^{6, 15}